The purpose of the current proposal is to use computational neuroscience to empirically identify endophenotypes of ASD in a large scale resting-state fMRI (rs-fMRI) repository (the Autism Brain Imaging Exchange [ABIDE], N~1,112 [ASD N=539]). Specifically, we propose to use resting-state connectivity maps in conjunction with group iterative multiple model estimation (Gates &Molenaar, 2012) and community structure detection (Newman, 2006) to generate empirically-derived subgroups of ASD who share similar brain network properties. The rationale for this approach is as follows. It has been widely acknowledged that ASD is a vastly heterogeneous disorder (e.g. Volkmar et al., 2004). It is also increasingly accepted that ASD is a "network-disorder", involving complex degradation of brain networks. However, prior work in network-analysis of ASD has generally ignored this heterogeneity, and proceeded in traditional between-groups (ASD vs. Control) comparison. Our objective here is to determine if heterogeneity within ASD can actually be useful information, which facilitates the identification of subgroups (communities) whose brain network properties are similar [AIM 1], and whose symptoms cluster together [AIM 2]. Our target network will be the Default-Mode Network (DMN), a brain system that is 1) anchored in the posteromedial cortex, and 2) involved in multiple forms of social cognition that are known to be disrupted in ASD. Based on recent, well- conducted studies of DMN in autism, (e.g., Lynch et al., 2013;Rudie et al., 2012;Washington et al., 2013), we hypothesize that DMN has a heterogeneous connectivity profile in ASD, and that connectivity within the DMN can be used to parse subtypes of autism. We further predict that these individual patterns of connectivity are strongly related to individua differences in the phenotypic presentation of ASD based on the topography of connections. Our approach represents a substantially different way of using heterogeneity, and we provide extensive simulations to demonstrate the computational feasibility of endophenotype generation, and also the method by which endophenotypes will be linked to behavioral data available through ABIDE. We feel that our novel approach, in combination with the largest ASD resting-state fMRI repository ever created will stimulate vertical progress by overcoming problems associated with small sample size, univariate approaches and missing or inconsistent phenotypic data.
Autism spectrum disorders (ASDs) are an increasingly common and disabling neurodevelopmental disorder. The proposed study is the first known attempt to use an extremely large sample of brain imaging data to identify whether subgroups of ASD exist. If subtypes of ASD can be found using brain imaging data, this would reveal important information about how symptoms are related to brain function, potentially leading to rational treatment approaches.
|White, Susan W; Richey, John A; Gracanin, Denis et al. (2015) The Promise of Neurotechnology in Clinical Translational Science. Clin Psychol Sci 3:797-815|