TBC
This work utilized behavior partial least squares correlation analysis in several datasets to derive multivariate brain-behavior associations between resting-state functional connectivity in relation to transdiagnostic psychiatric dimensions in middle-aged and older adults.
This repository contains code for analyzing multivariate brain�behavior relationships using Partial Least Squares Correlation (PLSC), along with downstream validation analyses including:
-
Longitudinal prediction of mental health outcomes
-
Mortality prediction using Cox proportional hazards models
-
External validation of brain scores
Requirements: Matlab v2018b, R v4.4.1
plsc_analysis.m is used to carry out the behavior partial least squares correlation analysis.
Output figures: Brain salience maps, Behavior correlation plots, Network-level importance.
This script requires the following toolbox: PLS software
longitudinal_prediction.R is used to predict future mental health trajectory based on baseline brain scores.
Output figures: Boxplot of brain score distribution across diagnostic groups.
mortality_prediction_cox.R is used to fit Cox proportional hazards models and test whether brain scores predict survival outcomes.
Output figures: Hazard ratio plotfor all independent variables in the Cox models.
external_validation.m is used to compute projected brain scores using FC of an external dataset and salience maps from PLSC analysis.
Output figures: Correlation plot between the projected brain score and tested behavior.
Quick check:
Synthetic data generated from random number generator can be found in synthetic_data folder. If the scripts were run correctly, you should obtain figures as seen in the fig folder.
This project utilizes data from UK Biobank study. Guidelines to apply for data access can be found here.
Please contact Thuan Tinh Nguyen (Sophia) at thuantinh.nguyen@u.nus.edu and Helen Zhou at helen.zhou@nus.edu.sg