cpm: Connectome-based Predictive Modeling

In NetworkToolbox: Methods and Measures for Brain, Cognitive, and Psychometric Network Analysis

Description

Suite of functions for Connectome-based Predictive Modeling (CPM). See and cite Finn et al., 2015; Rosenberg et al., 2016; Shen et al., 2017

• cpmIV

  Internal Validation method (Rosenberg et al., 2016; Shen et al., 2017). Using a leave-one-out approach, this method correlates a behavioral statistic bstat with each edge of a whole-brain network across participants. Using the significant edges in the network thresh, a connectome model is built (without the participant's network). A linear regression model is fit, with the behavioral statistic being regressed on the connectome model. The left out participants connectome model is then used with the linear regression weights to compute their predicted behavioral score. This is repeated for every participant. The predicted scores are correlated with their observed score. Significant values suggest that the connectome is related to the behavioral statistic

• cpmIVperm

  Performs a permutation test of the results obtained by cpmIV. The permutation test quantifies whether the results obtained by the original cpmIV are significantly different than a random model (see Shen et al., 2017)

• cpmEV

  UNDER DEVELOPMENT. External Validation method (Beaty et al., 2018). Performs similar function as cpmIV but uses data to train train_na the connectome model using a behavioral statistic train_b. This training connectome model is then used to predict another dataset valid_na, using the same behavioral statistic valid_b. The full training dataset FALSE or the leave-one-out overlap = TRUE approach can be used

• cpmFP

  Fingerprinting method (Finn et al., 2015). Uses CPM approach to identify participants across two sessions

• cpmFPperm

  Fingerprinting method (Finn et al., 2015). Uses permutation method to estimate the significance of of the cpmFP results

• cpmPlot

  Plots the CPM results

Usage

cpmIV(neuralarray, bstat, kfolds, covar, thresh = .01,
      connections = c("separate", "overall"), groups = NULL,
      method = c("mean", "sum"), model = c("linear","quadratic","cubic"),
      corr = c("pearson","spearman"), nEdges, 
      standardize = FALSE, cores, progBar = TRUE, plots = TRUE)
      
cpmIVperm(iter = 1000, ...)
      
cpmEV(train_na, train_b, valid_na, valid_b, thresh = .01,
      overlap = FALSE, progBar = TRUE)
      
cpmFP(session1, session2, progBar = TRUE)

cpmFPperm(session1, session2, iter = 1000, progBar = TRUE)

cpmPlot(cpm.obj, visual.nets = FALSE)

Arguments

neuralarray	Array from convertConnBrainMat function
bstat	Behavioral statistic for each participant with neural data (a vector)
kfolds	Numeric. Number of k-fold validation samples. Defaults to the number of participants in the sample (i.e., n), which is also known as leave-one-out validation. Recommended folds are 5 and 10
covar	Covariates to be included in predicting relevant edges (time consuming). Must be input as a list() (see examples)
thresh	Sets an α threshold for edge weights to be retained. Defaults to .01
connections	Character. Should positive and negative correlations be separated or used together? Defaults to "separate"
groups	Allows grouping variables to be used for plotting points. Must be a vector. Defaults to NULL
method	Use "mean" or "sum" of edge strengths in the positive and negative connectomes. Defaults to "mean"
model	Regression model to use for fitting the data. Defaults to "linear"
corr	Correlation method for assessing the relationship between the behavioral measure and edges between ROIs. Defaults to "pearson". Set to "spearman" for non-linear or monotonic associations
nEdges	Number of participants that are required to have an edge to appear in the plots. Defaults to 10 percent of edges in participants
standardize	Should the behavioral statistic (bstat) be standardized? Defaults to FALSE
cores	Number of computer processing cores to use when performing covariate analyses. Defaults to n - 1 total number of cores. Set to any number between 1 and maximum amount of cores on your computer
progBar	Should progress bar be displayed? Defaults to TRUE. Set to FALSE for no progress bar
plots	Should plots be plotted? Defaults to TRUE. Set to FALSE to hide plots
train_na	Training dataset (an array from convertConnBrainMat function)
train_b	Behavioral statistic for each participant for the training neural data (a vector)
valid_na	Validation dataset (an array from convertConnBrainMat function)
valid_b	Behavioral statistic for each participant for the validation neural data (a vector)
overlap	Should leave-one-out cross-validation be used? Defaults to FALSE (use full dataset, no leave-one-out). Set to TRUE to select edges that appear in every leave-one-out cross-validation network (time consuming)
session1	Array from convertConnBrainMat function (first session)
session2	Array from convertConnBrainMat function (second session)
iter	Number of iterations to perform. Defaults to 1000
cpm.obj	cpm object
visual.nets	Boolean. Uses qgraph to plot connectivity between the networks as a network. Defaults to FALSE. Set to TRUE to visualize the networks
...	Additional arguments to be passed from a cpm function

Value

cpmIV and cpmEV:

Returns a list containing:

results	A matrix containing: r coefficient (r), p-value (p-value), mean absolute error (mae), root mean square error (rmse)
posMask	Positive connectivity for input in BioImage Suite Connectivity Viewer
negMask	Negative connectivity for input in BioImage Suite Connectivity Viewer

cpmIVperm:

Returns a matrix containing p-values for positive and negative prediction models

cpmFP:

Returns a matrix containing the percentage and number of correctly identified subjects for sessions 1 and 2

cpmPlot:

Returns plot of connectivity differences between the positive and negative masks

Author(s)

Alexander Christensen <[e-mail]>

References

Beaty, R. E., Kenett, Y. N., Christensen, A. P., Rosenberg, M. D., Benedek, M., Chen, Q., Fink, A., Qiu, J., Kwapil, T. R., Kane, M. J., & Silvia, P. J. (2018). Robust prediction of individual creative ability from brain functional connectivity. Proceedings of the National Academy of Sciences, 115, 1087-1092.

Finn, E. S., Shen, X., Scheinost, D., Rosenberg, M. D., Huang, J., Chun, M. M., Papademetris, X., Constable, R. T. (2015). Functional connectome fingerprinting: Identifying individuals using patterns of brain connectivity. Nature Neuroscience, 18, 1664-1671.

Rosenberg, M. D., Finn, E. S., Scheinost, D., Papademetris, X., Shen, X., Constable, R. T., Chun, M. M. (2016). A neuromarker of sustained attention from whole-brain functional connectivity. Nature Neuroscience, 19, 165-171.

Shen, X. Finn, E. S., Scheinost, D., Rosenberg, M. D., Chun, M. M., Papademetris, X., Constable, R. T. (2017). Using connectome-based predictive modeling to predict individual behavior from brain connectivity. Nature Protocols, 12, 506-518.

Wei, T. & Simko, V.(2017). R package "corrplot": Visualization of a correlation matrix (Version 0.84).

Examples

# Load data
behav <- behavOpen

## Not run: 

# Create path to temporary file
temp <- tempfile()

# Download to temporary file
googledrive::drive_download(
paste("https://drive.google.com/file/d/",
"1T7_mComB6HPxJxZZwwsLLSYHXsOuvOBt",
"/view?usp=sharing", sep = ""),
path = temp
)

# Load resting state brain data
load(temp)

# Run cpmIV
res <- cpmIV(neuralarray = restOpen, bstat = behav, cores = 4)

# Plot cpmIV results
cpmPlot(res)


## End(Not run)

NetworkToolbox documentation built on May 28, 2021, 5:11 p.m.