MEA-NAP Outputs¶
Here is a list of the figures produced by MEA-NAP including sample figure legends. You can navigate to the different subfolders using these links:
Step 2A – Individual Neuronal Analysis (includes a subfolder for each MEA recording)
- Step 4 – Network Activity
- Step 4B – Network Activity – Group Comparisons
Step 1 - Spike Detection¶
Step 1A - Spike Detection Files¶
This folder contains the MATLAB file for each MEA recording with the spike times detected by each spike detection method and parameter selected. This folder can be used by MEA-NAP to perform Steps 1B – Step 5 without repeating the spike detection.
Each matlab file contains the following variables:
Channels: a vector containing the numeric identified for each channel
spikeDetectionResult: a structure containing the parameters used for spike detection (e.g., sampling rate)
spikeTimes: a cell with an entry per channel, each cell entry contains a structure where the field names are the spike detection method and the field entries are the spike times in seconds
spikeWaveforms: same format as spikeTimes, but where each field entry are the spike waveforms detected
Thresholds: same format as spikeTimes, but where each field entry are the absolute values of the threshold used for spike detection (in mV), they are NaN values for wavelet detection methods as they do not rely on a threshold
Step 1B – Spike Detection Checks¶
This folder contains subfolders for each Group. Each subfolder contains folders for each MEA recording by filename.
Subfolder – By Filename¶
Figure 1. Example Traces. Sample 60-millisecond-long filtered voltage traces from 9 electrodes (if the default number of electrodes to plot is selected in MEA-NAP) centered on at least one action potential. The MEA recording filename is at the top of the figure. The electrode number and time in the recording are shown above each voltage trace. The colored arrows indicate where one or more spike detection methods and parameters identified an action potential. Legend, bior1.3, bior1.5, and db2 are MATLAB wavelets used for template-based spike detection with the continuous wavelet transform. The median absolute deviation used for the threshold method is indicated with the prefix “thr.” This figure facilitates comparing the performance of the spike detection methods and parameters at the individual electrode level.
Figure 2. Spike Frequencies. Line graphs show the running spike frequency (binned by 1 second) detected by each spike detection method (line color) during the length of recording in minutes. Legend, bior1.3, bior1.5, and db2 are MATLAB wavelets used for template-based spike detection with the continuous wavelet transform. The median absolute deviation used for the threshold method is indicated with the prefix “thr.” This figure facilitates comparing the performance of the spike detection methods and parameters at the electrode level. The figure title (top) indicates the MEA recording filename. This figure facilitates comparing the performance of the spike detection methods and parameters.
Figure 3. Waveforms. Sample of 50 action potential action potential waveforms (gray) overlayed and mean waveform (black line) detected with each spike detection method (panel title) from a sample electrode (indicated above the panels). The figure title is the MEA recording filename. For some electrodes, if fewer than 50 action potentials were detected with a specific spike detection method/parameter, then all of the action potential waveforms will be overlayed. This figure facilitates comparing the performance of the spike detection methods and parameters.
Step 2 – Neuronal Activity¶
This folder contains two subfolders, one to evaluate the individual recordings and one with comparisons by age and group. The Step 2A folder contains subfolders for each group and, within each group folder, individual subfolders for each recording. The Step 2B folder contains 4 subfolders that compare metrics by node (electrode) and recording for age and group.
Step 2A – Individual Neuronal Analysis¶
Figure 1. Firing Rate By Electrode. Scatter plot (gray circles), mean ± s.e.m. (black circle with error bars) and density curve show the mean firing rate (MFR) for each electrode in the microelectrode array (MEA) recording in spikes per second (Hz). The MFR was calculated by dividing the number of action potentials detected by the length of the recording in seconds. Figure title (top) is the name of the recording.
Figure 2. Heatmap. Mean firing rate (MFR) by electrode (circles) in the spatial arrangement of the microelectrode array (MEA) in hertz (Hz, color bar) scaled to the recording (left) and to the entire dataset (right). In the left panel, differences in MFR can be observed between electrodes. In the right panel, the MFR can be seen in comparison to the whole dataset. Figure title (top) is the name of the recording and scaling relative to the range of MFR in the recording versus the entire dataset. For Axion Biosystems users, please use MEA-NAP version 1.10.2 or later. There was an error previously in the orientation of MEA grid in this plot (flipped on the diagonal), which has now been corrected.
Figure 3. Raster. Raster plots show mean firing rate (MFR) in hertz (Hz, color bar) in 1-second bins for each electrode (row) over the length of the recording (time, minutes) scaled to the range of MFR in the recording (top panel) and in the entire dataset (bottom panel). The MFR was calculated as the number of action potentials per second. In the top raster plot, differences in MFR can be observed between electrodes. In the bottom raster plot, the relative MFR can be seen in comparison to the whole dataset. Figure title (above each raster plot) is the name of the recording and scaling relative to the range of MFR in the recording versus the entire dataset.
Step 2B - Group Comparison¶
Subfolder 1 – Node by Group¶
Figure 1. Mean Firing Rate by Node. Scatter plots, mean (black circles with error bars), and density curves show the mean firing rate (MFR) for each electrode (colored circles) from all of the recordings in the dataset in hertz (Hz) by age for Group 1 (left panel), Group 2 (right panel). Error bars may not be visible where they are smaller than the size of the circle representing the mean. The MFR was calculated by dividing the number of action potentials detected divided by the length of the recording in seconds. Panel title (top) is the group name.
Subfolder 2 – Node by Age¶
Figure 1. Mean Firing Rate by Node. Scatter plots, mean (black circles with error bars), and density curves show the mean firing rate (MFR) for each electrode (colored circles) from all of the recordings in the dataset in hertz (Hz) by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The MFR was calculated by dividing the number of action potentials detected divided by the length of the recording in seconds. Panel title (top) is the age.
Subfolder 3 – Recordings by Group¶
Figure 1. Number of Active Electrodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of active electrodes (colored circles) for each recording in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. An active electrode is defined as MFR greater than 0.01 Hz (if default settings in MEA-NAP were used). Panel title (top) is the group name.
Figure 2. Mean Firing Rate. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean firing rate for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 3. Median Firing Rate. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the median firing rate for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 4. Network Burst Rate. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the network burst rate (per minute) for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 5. Mean Number of Electrodes Involved in Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean number of electrodes (channels) involved in network bursts for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 6. Mean Network Burst Length. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean length of network bursts (in seconds) for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 7. Mean Inter-spike Interval (ISI) within Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean ISI between action potentials within network bursts (in milliseconds) for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 8. Mean Inter-spike Interval (ISI) Outside of Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean ISI between network bursts (in milliseconds) for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 9. Coefficient of Variation in the Inter-network-burst Intervals (IBI). Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the coefficient of variation of the intervals (in milliseconds) between network bursts for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 10. Fraction of In Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the fraction of bursts that are occurring within network bursts for each recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left-panel panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Subfolder 4 – Recordings by Age¶
Figure 1. Number of Active Electrodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of active electrodes (colored circles) for each recording in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. An active electrode is defined as MFR greater than 0.01 Hz (if default settings in MEA-NAP were used). Panel title (top) is the age name.
Figure 2. Mean Firing Rate. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean firing rate for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Figure 3. Median Firing Rate. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the median firing rate for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Figure 4. Network Burst Rate. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the network burst rate (per minute) for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Figure 5. Mean Number of Electrodes Involved in Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean number of electrodes (channels) involved in network bursts for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Figure 6. Mean Network Burst Length. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean length of network bursts (in seconds) for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name. mean.
Figure 7. Mean Inter-spike Interval (ISI) within Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean ISI between action potentials within network bursts (in milliseconds) for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Figure 8. Mean Inter-spike Interval (ISI) Outside of Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean ISI between network bursts (in milliseconds) for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Figure 9. Coefficient of Variation in the Inter-network-burst Intervals (IBI). Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the coefficient of variation of the intervals (in milliseconds) between network bursts for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Figure 10. Fraction of In Network Bursts. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the fraction of bursts that are occurring within network bursts for each recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age name.
Step 3 – Functional Connectivity Edge Thresholding Check¶
Figure 1. Edge Thresholding Check for Probabilistic Thresholding. Top panel, line graphs for the average (green) and coefficient of variation (black) for the threshold value for significant functional connections (edges) as the number of repeats (iterations of circular shifts used to determine the threshold for significance edges) increases for an example MEA recording from the dataset. The filename includes the recording name and spike time tiling coefficient lag. The green line represents the mean and the green shaded area the standard deviation. Middle panel, Threshold values for a sample of the individual edges (black lines) as the number of repeats increases. Most threshold values stabilize between 120-180 iterations of the circular shifts. Bottom panel, adjacency matrices show the edges that are eliminated (below the threshold for a significant edge) as the number of repeats increases.
Step 4 – Network Activity¶
This folder contains two subfolders, one to evaluate the individual recordings and one with comparisons by age and group. The Step 4A folder contains individual subfolders for each recording. The Step 4B folder contains 7 subfolders that compare metrics by node (electrode) and recording for age and group.
Step 4A - Individual Network Analysis¶
This folder contains subfolder(s) for each group. Within each group folder, there are subfolders for each recording from that group. Within each recording folder, there are subfolders for the network plots for each spike time tiling coefficient (STTC) lag used to determine the functional connectivity and two figures.
Subfolders – Individual Recordings by Filename¶
Figure 1. Non-negative Matrix Factorization (NMF) Reveals Patterns of Activity in the Microelectrode Array (MEA) Recording. Top left, Raster plot of action potentials (black lines) by electrode (rows) over the length of the MEA recording in seconds. Top right – bottom right panels, Raster plots of action potentials in the top 3 components determined by non-negative matrix factorization (NMF). Middle right panel, proportion of variance explained as the number of NMF components increases. The dashed gray line indicates the number of NMF components that are sufficient to explain 95% of the neuronal activity in the MEA recording. Lower right panel, The mean square root residual as the number of NMF components increases for the MEA recording (observed) and the action potentials shuffled in the recording (random). The intersection (dashed gray line) indicates the number of significant NMF components.
Figure 2. Node Cartography Proportions. Diagram (top left panel) shows how node cartography roles (colored circles, legend on bottom left panel) are determined using the within-module degree z-score and participation coefficient for each node. The boundaries (solid and dashed lines) between roles are automatically set based on the distribution in the entire dataset. Network schema (bottom left panel) illustrates node cartography roles. Bar graphs (right panel) compare proportion of nodes in each node cartography role (color) by spike time tiling coefficient (STTC) lag used to infer functional connectivity. Title of the figure is the MEA recording filename.
Subfolders – Network Activity By Lag (in milliseconds)¶
For each spike time tiling coefficient (STTC) lag used to determine the functional connections (edges), there is a separate folder for the network activity outputs of the individual MEA recordings. For Axion Biosystems users, please use MEA-NAP version 1.10.2 or later. There was an error previously in the orientation of MEA grid in this plot (flipped on the diagonal), which has now been corrected. This effects the orientation of the following figures in this subfolder: 2-5, 9, 10-11.
Figure 1. Adjacency Matrix and Functional Connectivity Statistics. Top left, Adjacency matrix shows significant edges and edge weights for the functional connections between individual nodes (neuronal activity from neuron or neurons at each electrode). The correlation coefficient was determined using the spike time tiling coefficient (STTC) with a lag (in milliseconds) indicated in the filename. Bottom left, Bar graphs show the maximum and mean correlation values for edges in the MEA recording. Top right, Histogram of node degree (number of significant connections) for nodes (electrodes) participating in the network activity. Middle right, Histogram of node strength (sum of the edge weights for each node). Bottom right, Histogram of significant edge weights (strength of function connections).
Figure 2. MEA Network Plot. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The size of the nodes and thickness of the edges are scaled based on the distributions in this recording.
Figure 2. Scaled MEA Network Plot. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The size of the nodes and thickness of the edges are scaled based on the distributions in the entire dataset to facilitate comparisons between MEA recordings.
Figure 2. Combined MEA Network Plots. Graphs of the functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The size of the nodes and thickness of the edges are scaled based on the distributions in this MEA recording (left) and the entire dataset (right) to facilitate comparison of the variation within the MEA recording and relative to other MEA recordings in the dataset.
Figure 3. MEA Network Plot with the Betweenness Centrality. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the betweenness centrality, a metric of what proportion of shortest paths, between any two nodes in the network, go through that node. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The betweenness centrality color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this recording.
Figure 3. Scaled MEA Network Plot with the Betweenness Centrality. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the betweenness centrality, a metric of what proportion of shortest paths, between any two nodes in the network, go through that node. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The betweenness centrality color bar, size of the nodes and thickness of the edges are scaled based on the distributions in the entire dataset to facilitate comparisons between MEA recordings.
Figure 3. Combined MEA Network Plots with the Betweenness Centrality. Graphs of the functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the betweenness centrality, a metric of what proportion of shortest paths, between any two nodes in the network, go through that node. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The betweenness centrality color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this MEA recording (left) and the entire dataset (right) to facilitate comparison of the variation within the MEA recording and relative to other MEA recordings in the dataset.
Figure 4. MEA Network Plot with the Participation Coefficient. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the participation coefficient, a metric of how well distributed a node’s edges are among different modules in the network. Values near 0 indicate the node’s edges are restricted to other nodes in the same module, while values near 1 indicate the node’s edges are evenly distributed among modules. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The participation coefficient color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this recording.
Figure 4. Scaled MEA Network Plot with the Participation Coefficient. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the participation coefficient, a metric of how well distributed a node’s edges are among different modules in the network. Values near 0 indicate the node’s edges are restricted to other nodes in the same module, while values near 1 indicate the node’s edges are evenly distributed among modules. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The participation coefficient color bar, size of the nodes and thickness of the edges are scaled based on the distributions in the entire dataset to facilitate comparisons between MEA recordings.
Figure 4. Combined MEA Network Plots with the Participation Coefficient. Graphs of the functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the participation coefficient, a metric of how well distributed a node’s edges are among different modules in the network. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The participation coefficient color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this MEA recording (left) and the entire dataset (right) to facilitate comparison of the variation within the MEA recording and relative to other MEA recordings in the dataset.
Figure 5. MEA Network Plot with the Local Efficiency. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the local efficiency, a metric of how well the node is connected to its nearest neighbors. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The local efficiency color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this recording.
Figure 5. Scaled MEA Network Plot with the Local Efficiency. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the local efficiency, a metric of how well the node is connected to its nearest neighbors. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The local efficiency color bar, size of the nodes and thickness of the edges are scaled based on the distributions in the entire dataset to facilitate comparisons between MEA recordings.
Figure 5. Combined MEA Network Plots with the Local Efficiency. Graphs of the functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the local efficiency, a metric of how well the node is connected to its nearest neighbors. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The local efficiency color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this MEA recording (left) and the entire dataset (right) to facilitate comparison of the variation within the MEA recording and relative to other MEA recordings in the dataset.
Figure 6. Circular Network Plot with Modules. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode arranged by module (subcommunities within the network). Nodes with the same color are part of the same module. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity.
Figure 7. Graph Theoretical Metrics By Node. Summary plots of nodal- and edge-level graph theoretical metrics for the MEA recording. Top row, diagram of network metrics. Bottom row, Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves for node degree, edge weight, node strength, within-module degree z-score, local efficiency, participation coefficient, and betweenness centrality. These graph metrics were calculated from the adjacency matrix for the MEA recording using the spike time tiling coefficient (STTC) lag indicated in the title.
Figure 8. Null Models for Small-World Coefficient (ω). Line graphs show the small-world coefficient (blue lines) for a lattice (top) and random (bottom) network as the number of iterations of circular shifts of the activity in the original MEA recording increases. This plot is used to check that the number of iterations was sufficient for creating the null models used to normalize the small-world coefficient (ω). The MEA recording filename and spike time tiling coefficient (STTC) lag are indicated in the title.
Figure 9. Circular Node Cartography Network Plot. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode arranged by module (subcommunities within the network). The node color indicates the node cartography role. Gray circles with no edges (when present) indicate electrodes without neurons participating in the network activity. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity.
Figure 9. MEA Network Plot with the Node Cartography. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the node cartography role. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity.
Figure 9. Node Cartography for Adjacency Matrix by STTC Lag. Diagram (top panel) shows how the node cartography role for each node (colored circles, legend on bottom left panel) are determined using the within-module degree z-score and participation coefficient for each node. The boundaries (dashed lines) between roles were automatically set based on the distribution in the entire dataset. Network schema (bottom right panel) illustrates node cartography roles. Title of the figure is the MEA recording filename and the spike time tiling coefficient (STTC) lag used to create the adjacency matrix.
Figure 10. MEA Network Plot with the Average Controllability. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the average controllability, a metric of how much influence a node has over the overall network activity. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The average controllability color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this recording.
Figure 10. Scaled MEA Network Plot with the Average Controllability. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the average controllability, a metric of how much influence a node has over the overall network activity. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The average controllability color bar, size of the nodes and thickness of the edges are scaled to the theoretical maximum and minimum to facilitate comparisons between MEA recordings.
Figure 10. Combined MEA Network Plots with the Average Controllability. Graphs of the functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the average controllability, a metric of how much influence a node has over the overall network activity. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The average controllability color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this MEA recording (left) and the theoretical maximum and minimum (right) to facilitate comparison of the variation within the MEA recording and relative to other MEA recordings in the dataset.
Figure 11. MEA Network Plot with the Modal Controllability. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the modal controllability. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The modal controllability color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this recording.
Figure 11. Scaled MEA Network Plot with Modal Controllability. Graph of functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the modal controllability. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The modal controllability color bar, size of the nodes and thickness of the edges are scaled to the theoretical maximum and minimum to facilitate comparisons between MEA recordings.
Figure 11. Combined MEA Network Plots with the Modal Controllability. Graphs of the functional connectivity for an individual MEA recording (filename and STTC lag indicated in title). The nodes (circles) represent the neuronal activity observed from neuron(s) at each electrode in the spatial arrangement of the MEA. The node color represents the modal controllability. The node degree (size of circle) represents the number of functional connections with other nodes. The edges (lines) represent significant functional connections between nodes, and the edge weight (line thickness) represents the strength of connectivity. The modal controllability color bar, size of the nodes and thickness of the edges are scaled based on the distributions in this MEA recording (left) and the theoretical maximum and minimum (right) to facilitate comparison of the variation within the MEA recording and relative to other MEA recordings in the dataset.
Step 4B – Network Activity – Group Comparisons¶
Subfolder 1 – Node by Group (with subfolders for each STTC lag)¶
Figure 1. Node Degree by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the node degree for node (colored circles) for all of the recordings in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The node degree was calculated as the number of significant edges for each node. Panel title (top) is the group name.
Figure 2. Edge Weight by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the edge weights for all edges (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The edge weights were calculated using spike time tiling coefficient (STTC). Panel title (top) is the group name.
Figure 3. Node Strength by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the node strength for all nodes (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The node strength is the sum of the edge weights for each node’s connections. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The edge weights were calculated using spike time tiling coefficient (STTC). Panel title (top) is the group name.
Figure 4. Local efficiency by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the local efficiency for all nodes (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 5. Within-module Degree z-Score by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the within-module degree z-score for all nodes (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 6. Betweenness Centrality by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the betweenness centrality for all nodes (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 7. Participation Coefficient by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the participation coefficient for all nodes (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Values near 0 indicate the node’s edges are restricted to other nodes in the same module, while values near 1 indicate the node’s edges are evenly distributed among modules. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 8. Average Controllability by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the average controllability for all nodes (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 9. Modal controllability by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the modal controllability for all nodes (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Subfolder 2 – Node by Age (with subfolders for each STTC lag)¶
Figure 1. Node Degree by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the node degree for node (colored circles) for all of the recordings in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The node degree was calculated as the number of significant edges for each node. Panel title (top) is the age.
Figure 2. Edge Weight by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the edge weights for all edges (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The edge weights were calculated using spike time tiling coefficient (STTC). Panel title (top) is the age.
Figure 3. Node Strength by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the node strength for all nodes (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The node strength is the sum of the edge weights for each node’s connections. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The edge weights were calculated using spike time tiling coefficient (STTC). Panel title (top) is the age.
Figure 4. Local efficiency by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the local effeciency for all nodes (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 5. Within-module Degree z-Score by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the within-module degree z-score for all nodes (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 6. Betweenness Centrality by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the betweenness centrality for all nodes (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 7. Participation Coefficient by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the participation coefficient for all nodes (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Values near 0 indicate the node’s edges are restricted to other nodes in the same module, while values near 1 indicate the node’s edges are evenly distributed among modules. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 8. Average Controllability by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the average controllability for all nodes (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 9. Modal controllability by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the modal controllability for all nodes (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Subfolder 3 – Recordings by Group (with subfolders for each STTC lag)¶
Figure 1. Network Size by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the network size for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The network size was calculated as number of nodes with at least one significant edge. Panel title (top) is the group name.
Figure 2. Network Density by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the network density for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The density was calculated as proportion of significant edges as a function of the total possible edges. Panel title (top) is the group name.
Figure 3. Mean Node Degree by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean node degree per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The node degree is calculated for each node in the network as the number of significant connections with other nodes and were averaged for each recording. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 4. Mean Node Degree of the Top 25% of Nodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean node degree for the top 25% of nodes per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The node degree is calculated for each node in the network as the number of significant connections with other nodes and the top 25% of nodes’ node degrees were averaged for each recording. This metric enables comparison of the most highly connected nodes in the networks. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 5. Mean Significant Edge Weight by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean of the significant edge weights for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Significant edges and their weight are determined using the spike time tiling coefficient and probabilistic thresholding. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 6. Mean Edge Weight of the Top 10% of Significant Edges. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean edge weight for the top 10% of edges per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. This metric enables comparison of the strongest significant edges (most highly correlated activity) in the networks. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 7. Mean Node Strength by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean node strength for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Node strength is calculated for each node as the sum of its edge weights and then averaged for all nodes in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 8. Mean Local Efficiency by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean local efficiency for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The local efficiency is calculated for each node and then averaged for all nodes in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 9. Clustering Coefficient by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the clustering coefficient for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 10. Number of Modules by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of modules for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 11. Modularity Score by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the modularity score for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 12. Percentage of Nodes with Within-module Degree Z-scores Greater than Zero. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the percentage of nodes with within-module degree z-scores greater than zero for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The within-module degree z-score is a nodal-level metric for which higher values indicate more intermodular connections (e.g., as seen in hub nodes). Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 13. Percentage of Nodes with Within-module Degree Z-scores Less than Zero. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the percentage of nodes with within-module degree z-scores less than zero for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The within-module degree z-score is a nodal-level metric for which lower values indicate fewer intramodular connections (e.g., non-hub or peripheral nodes). Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 14. Mean Path Length by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean path length for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 15. Mean Participation Coefficient by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean participation coefficient for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The participation coefficient is calculated for each node and then averaged for all nodes in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 16. Mean Participation Coefficient of the Bottom 10% of Nodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean participation coefficient for the bottom 10% of nodes per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The participation coefficient is calculated for each node in the network and the bottom 10% of nodes’ participation coefficients were averaged for each recording. This metric may be particular information in highly connected networks to compare nodes with higher modularity that are not participating in highly correlated network activity. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 17. Mean Participation Coefficient of the Top 10% of Nodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean participation coefficient for the top 10% of nodes per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The participation coefficient is calculated for each node in the network and the top 10% of nodes’ participation coefficients were averaged for each recording. This metric enables comparison of the nodes with edges that are evenly distributed among modules in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 18. Global Efficiency by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the global efficiency for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 19. Proportion of Peripeheral Nodes by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of peripheral nodes for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 20. Proportion of Non-hub Connectors by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of non-hub connector nodes for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 21. Proportion Non-hub Kinless Nodes by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of non-hub kinless nodes for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 22. Proportion of Provincial Hubs by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of provincial hubs for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 23. Proportion of Connector Hubs by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of connector hubs for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 24. Proportion of Kinless Hubs by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of kinless hubs for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 25. Small-world Coefficient (σ) by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the small-world coefficient (σ) for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The small-world coefficient (σ) is calculated as clustering coefficient divided by characteristic path length. Small-world networks have a value of σ >1. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 26. Small-world Coefficient (ω) by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the small-world coefficient (ω) for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The small-world coefficient (ω) is calculated using the normalized clustering coefficient and path length. Small-world network structure is at the midpoint (0) between a lattice (-1) and random (1) network structure. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 27. Mean Average Controllability by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean average controllability for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The average controllability is a measure of how much influence a node has over the overall network activity. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 28. Number of Significant Non-negative Matrix Factorization (NMF) Components by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of significant NMF components for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. NMF identifies patterns of network activity within the network in MEA recordings. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 29. Number of Significant Non-negative Matrix Factorization (NMF) Components Divided by Network Size. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of significant NMF components divided by network size for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Normalizing the number of significant NMF components by network size can facilitate comparison between networks. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 30. Effective Rank by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the effective rank for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. Effective rank is a measure of the number of subcommunities in the network based on network activity patterns. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Subfolder 4 – Recordings by Age (with subfolders for each STTC lag)¶
Figure 1. Network Size by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the network size for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The network size was calculated as number of nodes with at least one significant edge. Panel title (top) is the age.
Figure 2. Network Density by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the network density for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. The density was calculated as proportion of significant edges as a function of the total possible edges. Panel title (top) is the age.
Figure 3. Mean Node Degree by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean node degree for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The node degree is calculated for each node in the network as the number of significant connections with other nodes and the averaged for each recording. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 4. Mean Node Degree of the Top 25% of Nodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean node degree for the top 25% of nodes per MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The node degree is calculated for each node in the network as the number of significant connections with other nodes and the top 25% of nodes’ node degrees were averaged for each recording. This metric enables comparison of the most highly connected nodes in the networks. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 5. Mean Significant Edge Weight by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean of the significant edge weights for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Significant edges and their weight are determined using the spike time tiling coefficient and probabilistic thresholding. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 6. Mean Edge Weight of the Top 10% of Significant Edges. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean edge weight for the top 10% of edges per MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. This metric enables comparison of the strongest significant edges (most highly correlated activity) in the networks. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 7. Mean Node Strength by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean node strength for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Node strength is calculated for each node as the sum of its edge weights and then averaged for all nodes in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 8. Mean Local Efficiency by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean local efficiency for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The local efficiency is calculated for each node and then averaged for all nodes in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 9. Clustering Coefficient by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the clustering coefficient for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 10. Number of Modules by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of modules for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 11. Modularity Score by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the modularity score for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 12. Percentage of Nodes with Within-module Degree Z-scores Greater than Zero. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the percentage of nodes with within-module degree z-scores greater than zero for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The within-module degree z-score is a nodal-level metric for which higher values indicate more intermodular connections (e.g., as seen in hub nodes). Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 13. Percentage of Nodes with Within-module Degree Z-scores Less than Zero. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the percentage of nodes with within-module degree z-scores less than zero for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The within-module degree z-score is a nodal-level metric for which lower values indicate fewer intramodular connections (e.g., non-hub or peripheral nodes). Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 14. Mean Path Length by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean path length for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 15. Mean Participation Coefficient by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean participation coefficient for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The participation coefficient is calculated for each node and then averaged for all nodes in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 16. Mean Participation Coefficient of the Bottom 10% of Nodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean participation coefficient for the bottom 10% of nodes per MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The participation coefficient is calculated for each node in the network and the bottom 10% of nodes’ participation coefficients were averaged for each recording. This metric may be particular information in highly connected networks to compare nodes with higher modularity that are not participating in highly correlated network activity. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 17. Mean Participation Coefficient of the Top 10% of Nodes. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean participation coefficient for the top 10% of nodes per MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The participation coefficient is calculated for each node in the network and the top 10% of nodes’ participation coefficients were averaged for each recording. This metric enables comparison of the nodes with edges that are evenly distributed among modules in the network. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 18. Global Efficiency by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the global efficiency for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 19. Proportion of Peripeheral Nodes by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of peripheral nodes for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. calculated using node cartography. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 20. Proportion of Non-hub Connectors by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of non-hub connectors for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. calculated using node cartography. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 21. Proportion of Non-hub Kinless Nodes by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of non-hub connectors for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. calculated using node cartography. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 22. Proportion of Provincial Hubs by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of provincial hubs for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. calculated using node cartography. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 23. Proportion of Connector Hubs by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of connector hubs for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. calculated using node cartography. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 24. Proportion of Kinless Hubs by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the proportion of kinless hubs for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. calculated using node cartography. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 25. Small-world Coefficient (σ) by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the small-world coefficient (σ) for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. The small-world coefficient (σ) is calculated as clustering coefficient divided by characteristic path length. Small-world networks have a value of σ >1. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 26. Small-world Coefficient (ω) by Group. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the small-world coefficient (ω) for each MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The small-world coefficient (ω) is calculated using the normalized clustering coefficient and path length. Small-world network structure is at the midpoint (0) between a lattice (-1) and random (1) network structure. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the group name.
Figure 27. Mean Average Controllability by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the mean average controllability for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 28. Number of Significant Non-negative Matrix Factorization (NMF) Components by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of significant NMF components for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. NMF identifies patterns of network activity within the network in MEA recordings. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 29. Number of Significant Non-negative Matrix Factorization (NMF) Components Divided by Network Size. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the number of significant NMF components divided by network size for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Normalizing the number of significant NMF components by network size can facilitate comparison between networks. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Figure 30. Effective Rank by Age. Scatter plots, mean ± s.e.m. (black circles with error bars), and density curves show the effective rank for each MEA recording (colored circles) in the dataset by group for Age 1 (left panel), Age 2 (second-from-the-left-panel), etc. Effective rank is a measure of the number of subcommunities in the network based on network activity patterns. Error bars may not be visible where they are smaller than the size of the circle representing the mean. Panel title (top) is the age.
Subfolder 5 – Graph Metrics by Lag¶
Figure 1. Network Size by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) network size by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of choice of STTC lag on network size. For datasets with more than one group, each panel title indicates the group name.
Figure 2. Network Density by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) network density by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of choice of STTC lag on the network density. For datasets with more than one group, each panel title indicates the group name.
Figure 3. Mean Node Degree by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) mean node degree by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. The node degree is calculated for each node in the network as the number of significant connections with other nodes and the averaged for each recording. This figure illustrates the impact of the choice of STTC lag on the mean node degree. For datasets with more than one group, each panel title indicates the group name.
Figure 4. Mean Node Degree of the Top 25% of Nodes. Line graphs of the mean (solid line) ± s.e.m. (shading) mean node degree for the top 25% of nodes per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The node degree is calculated for each node in the network as the number of significant connections with other nodes and the top 25% of nodes’ node degrees were averaged for each recording. This metric enables comparison of the most highly connected nodes in the networks. This figure illustrates the impact of the choice of STTC lag on the mean node degree of the top 25% of nodes. For datasets with more than one group, each panel title indicates the group name.
Figure 5. Mean Significant Edge Weight by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) mean of the significant edge weights by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. Significant edges and their weight are determined using the STTC and probabilistic thresholding. This figure illustrates the impact of the choice of STTC lag on the mean of significant edge weights. For datasets with more than one group, each panel title indicates the group name.
Figure 6. Mean Edge Weight of the Top 10% of Significant Edges. Line graphs of the mean (solid line) ± s.e.m. (shading) mean edge weight for the top 10% of edges per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. This metric enables comparison of the strongest significant edges (most highly correlated activity) in the networks. For datasets with more than one group, each panel title indicates the group name.
Figure 7. Mean Node Strength by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) mean node strength by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. Node strength is calculated for each node as the sum of its edge weights and then averaged for all nodes in the network. This figure illustrates the impact of the choice of STTC lag on the mean node strength. For datasets with more than one group, each panel title indicates the group name.
Figure 8. Mean Local Efficiency by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) mean local efficiency by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. The local efficiency is calculated for each node and then averaged for all nodes in the network. This figure illustrates the impact of the choice of STTC lag on the mean local efficiency. For datasets with more than one group, each panel title indicates the group name.
Figure 9. Clustering Coefficient by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) clustering coefficient by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the clustering coefficient. For datasets with more than one group, each panel title indicates the group name.
Figure 10. Number of Modules by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) number of modules by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the number of modules. For datasets with more than one group, each panel title indicates the group name.
Figure 11. Modularity Score by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) modularity score by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the modularity score. For datasets with more than one group, each panel title indicates the group name.
Figure 12. Percentage of Nodes with Within-module Degree Z-scores Greater than Zero. Line graphs of the mean (solid line) ± s.e.m. (shading) percentage of nodes with within-module degree z-scores greater than zero by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. The within-module degree z-score is a nodal-level metric for which higher values indicate more intermodular connections (e.g., as seen in hub nodes). For datasets with more than one group, each panel title indicates the group name.
Figure 13. Percentage of Nodes with Within-module Degree Z-scores Less than Zero. Line graphs of the mean (solid line) ± s.e.m. (shading) percentage of nodes with within-module degree z-scores less than zero by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. The within-module degree z-score is a nodal-level metric for which lower values indicate fewer intramodular connections (e.g., non-hub or peripheral nodes). For datasets with more than one group, each panel title indicates the group name.
Figure 14. Mean Path Length by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) mean path length by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the mean path length. For datasets with more than one group, each panel title indicates the group name.
Figure 15. Mean Participation Coefficient by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) mean participation coefficient by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. The participation coefficient is calculated for each node and then averaged for all nodes in the network. This figure illustrates the impact of the choice of STTC lag on the mean participation coefficient. For datasets with more than one group, each panel title indicates the group name.
Figure 16. Mean Participation Coefficient of the Bottom 10% of Nodes. Line graphs of the mean (solid line) ± s.e.m. (shading) mean participation coefficient for the bottom 10% of nodes per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The participation coefficient is calculated for each node in the network and the bottom 10% of nodes’ participation coefficients were averaged for each recording. This metric may be particular information in highly connected networks to compare nodes with higher modularity that are not participating in highly correlated network activity. For datasets with more than one group, each panel title indicates the group name.
Figure 17. Mean Participation Coefficient of the Top 10% of Nodes. Line graphs of the mean (solid line) ± s.e.m. (shading) mean participation coefficient for the top 10% of nodes per MEA recording (colored circles) in the dataset by age for Group 1 (left panel), Group 2 (second-from-the-left panel), etc. The participation coefficient is calculated for each node in the network and the top 10% of nodes’ participation coefficients were averaged for each recording. This metric enables comparison of the nodes with edges that are evenly distributed among modules in the network. For datasets with more than one group, each panel title indicates the group name.
Figure 18. Global Efficiency by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) global efficiency by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the global efficiency. For datasets with more than one group, each panel title indicates the group name.
Figure 19. Proportion of Peripeheral Nodes by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) proportion of peripheral nodes by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the proportion of peripheral nodes. For datasets with more than one group, each panel title indicates the group name.
Figure 20. Proportion of Non-hub Connectors by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) proportion of non-hub connectors by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the proportion of non-hub connectors. For datasets with more than one group, each panel title indicates the group name.
Figure 21. Proportion of Non-hub Kinless Nodes by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) proportion of non-hub kinless nodes by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the proportion of non-hub kinless nodes. For datasets with more than one group, each panel title indicates the group name.
Figure 22. Proportion of Provincial Hubs by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) proportion of provincial hubs by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the proportion of provincial hubs. For datasets with more than one group, each panel title indicates the group name.
Figure 23. Proportion of Connector Hubs by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) proportion of connector hubs by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the proportion of connector hubs. For datasets with more than one group, each panel title indicates the group name.
Figure 24. Proportion of Kinless Hubs by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) proportion of kinless hubs by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the proportion of kinless hubs. For datasets with more than one group, each panel title indicates the group name.
Figure 25. Small-world Coefficient (σ) by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) small-world coefficient (σ) by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. The small-world coefficient (σ) is calculated as clustering coefficient divided by characteristic path length. Small-world networks have a value of σ >1. This figure illustrates the impact of the choice of STTC lag on the small-world coefficient. For datasets with more than one group, each panel title indicates the group name.
Figure 26. Small-world Coefficient (ω) by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) small-world coefficient (ω) by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. The small-world coefficient (ω) is calculated using the normalized clustering coefficient and path length. For small-world networks, ω is at the midpoint (0) between a lattice (-1). This figure illustrates the impact of the choice of STTC lag on the small-world coefficient. For datasets with more than one group, each panel title indicates the group name.
Figure 27. Mean Average Controllability by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) mean average controllability by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. The STTC and probabilistic thresholding are used to determine the significant edges in the network. This figure illustrates the impact of the choice of STTC lag on the mean average controllability. For datasets with more than one group, each panel title indicates the group name.
Figure 28. Number of Significant Non-negative Matrix Factorization (NMF) Components by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) number of significant NMF components by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. NMF identifies patterns of network activity within the network in MEA recordings. This figure illustrates the impact of the choice of STTC lag on the number of significant NMF components. For datasets with more than one group, each panel title indicates the group name.
Figure 29. Number of Significant Non-negative Matrix Factorization (NMF) Components Divided by Network Size by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) number of significant NMF components divided by network size by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. Normalizing the number of significant NMF components by network size can facilitate comparison between networks. For datasets with more than one group, each panel title indicates the group name.
Figure 30. Effective Rank by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) effective rank by age (colors) for different spike time tiling coefficient (STTC) lags (x-axis) in milliseconds. Effective rank is a measure of the number of subcommunities in the network based on network activity patterns. This figure illustrates the impact of the choice of STTC lag on the effective rank. For datasets with more than one group, each panel title indicates the group name.
Subfolder 6 – Node Cartography By Lag¶
There will be one figure per spike time tiling coefficient (STTC) lag selected when MEA-NAP was run. The STTC lag (in milliseconds) is indicated in the figure filename.
Figure 1. Node Cartography by Spike Time Tiling Coefficient (STTC) Lag. Line graphs of the mean (solid line) ± s.e.m. (shading) proportion of each node cartograph role (colors) by age (x-axis). Figure legend (right) indicates the color of the node cartography roles. This figure facilitates comparing the mean proportion of node cartography roles by age and group. For datasets with more than one group, each panel (arranged vertically top to bottom) indicates the group name in the title.
Subfolder 7 – Density Landscape¶
Figure 1. Density Landscape for Determining the Node Cartography. Scatterplot shows values of within-module degree z-score and participation coefficient for all of the nodes (blue circles) in the entire dataset. The colored lines show the automated k-means boundaries set for determining the hub and non-hub roles (horizontal gray line, based on within-module degree z-score) and for the node cartography roles within the hub and non-hub designations (vertical colored lines, based on the participation coefficient). This figure was created to evaluate the automated boundaries set for determining the node cartography roles.
Step 5 – Statistical Comparisons¶
There will be one figure per spike time tiling coefficient (STTC) lag selected when MEA-NAP was run. The STTC lag (in milliseconds) is indicated in the figure filename.
Subfolders - Stats Figures by Lag¶
Figure 1. Confusion Matrix for All Classifiers. Confusion matrices for each classifier show true number of samples placed into each age group versus predicted class, the number the classifier placed into each age group. Blue indicates a correct and red an incorrect prediction by the classifier. Classifiers compared include the linear support-vector machine (SVM), k-nearest neighbors (kNN) algorithm, decision tree, and linear discriminant analysis (LDA). Note, this figure is only generated if more than one time point is provided in the dataset.
Figure 2. Misclassification Rate per K Fold for All Classifiers. Scatter plot shows the misclassification rate for each classifier during k-fold cross validation (k is set to 5 by default). Lower values indicate better classification performance. The gray horizontal line depicts the misclassification rate at chance level. Classifiers compared include the linear support-vector machine (SVM), k-nearest neighbors (kNN) algorithm, decision tree, and linear discriminant analysis (LDA). Note, this figure is only generated if more than one time point is provided in the dataset.
Figure 3. Leave One Feature Out for All Classifiers. Line graphs show different between leave one out loss and the original loss for four classifiers (colored lines) for network metrics including number of active nodes (aN), network density (Dens), clustering coefficient (CC), number of modules (nMod), modularity score (Q), path length (PL), global efficiency (Eglob), small-worldness coefficient (σ) and (ω), effective rank and the number of significant NMF components. Higher values indicate better fit for the classifier. Classifiers compared include the linear support-vector machine (SVM), k-nearest neighbors (kNN) algorithm, decision tree, and linear discriminant analysis (LDA). Note, this figure is only generated if more than one time point is provided in the dataset.
Figure 4. Nodal-level Feature Correlation. Correlation matrices for each group and age combination for nodal-level network features including node degree (ND), mean edge weight (MEW), node strength (NS), within-module z-score (Z), local efficiency (Eloc), participation coefficient (PC), and betweenness centrality (BC). Color (scale bar) indicates correlation value (0-1).
Figure 5. Recording-level Feature Correlation. Correlation matrices for each group and age combination for recording-level network features. Color (scale bar) indicates correlation value (0-1).