Implementing best practices in software design for Network Level Analysis

点击翻译按钮获取中文摘要

PROJECT SUMMARY Contemporary research views the brain as a large-scale, complex network composed of nonadjacent, yet connected brain regions. Rather than focusing on a limited set of a priori regions of interest, the field of neuroimaging has shifted towards statistical testing on associations across the whole connectome, i.e., at every possible brain connection. However, these connectome-wide association studies have a severe multiple comparisons problem, necessitating statistical methods which can control the false positive rate for associations between behavior and upwards of 50k functional connections. The long-term goal of the Parent BRAIN Initiative R00 (EB029343, ‘Innovative biostatistical approaches to network level analyses of connectome-behavior relationships’) is to create a statistical analysis software that would leverage the inherent network architecture of the connectome in order to probe fundamental biological mechanisms underlying the development of healthy and disordered cognition, behavior, and emotion. Specifically, the parent grant aims to formalize and validate in house analysis pipelines into a Network Level Analysis (NLA) toolbox as a comprehensive, versatile tool for use in connectome-wide association studies. While the research focus of this career transition award is on the application of NLA to developmental mechanisms of executive function and emotion regulation, this versatile analytic tool will be transformative to connectome data analysis across species, across the lifespan, and in health and disease. As part of tool development during the K99/R00, Dr. Wheelock has validated multiple NLA approaches, establishing sensitivity and specificity of network level findings using in silico connectome-behavior relationships, test-retest reliability of NLA approaches using in vivo human connectome and behavioral data from the HCP-Young Adult cohort, and ongoing work is extending NLA to investigate changes in connectome architecture supporting the development of executive and emotional function using connectome and behavioral data from the ABCD study (N=11,000 age 9-14). In Aim 2 of the R00, NLA toolbox is being updated to reflect object-oriented programming, incorporating longitudinal models and a graphical user interface. The goal of this Administrative Supplement is to improve NLA functionality by implementing several crucial changes. Specifically, funding from this Administrative Supplement, NOT-OD-23-073, will promote refactorization of NLA to improve computational efficiency, and usability by both developers and end users. The goals of this Supplement are to 1) refactor and optimize computational modeling in lower-level programming languages, 2) incorporate error logging and expand documentation, and 3) establish unit and integration testing to improve code merging. Successful completion of these Aims will both complement and extend the impact of the Parent R00, significantly improving the functionality and sustainability of NLA software in keeping with best practices of open science as well as increase accessibility of the software, enabling community-wide adoption of network-analysis methods for connectome-wide association studies.