Network Analysis of Protein Interaction Module of AutDB Database

Background:  Hundreds of genetic variants have been linked to autism, complicating the understanding of disease etiology. One approach to decode this heterogeneous genetic landscape is to identify shared molecular pathways of autism candidate genes that are known to often work together in a common biological context. In order to systematically catalogue interactions among autism genes, we have developed the protein interaction (PIN) module that integrates into our previously developed autism database (AutDB). 

Objectives:  The goal of this work is twofold: To (1) maintain a comprehensive, up-to-date integrated module (PIN)  comprising of systematically curated, manually verified, molecular interactions of autism genes (2) perform network analysis of interactions among autism-linked genes using PIN module.

Methods:  PIN module manual curation process entails data extraction from published, peer-reviewed literature. Post consultation with public databases (BioGRID, HPRD), PubMed searches are performed and three types of binding (protein, RNA, promoter) and three types of regulation (protein modification, direct regulation, autoregulation) interactions are manually extracted.  Utilizing this data in PIN module, we built networks for a reference gene set consisting of rare and syndromic ASD genes. Next, we identified all instances of direct molecular interactions among the autism risk genes in this network. To evaluate the statistical significance of these direct interactions, we compared the topological properties of the network with 1000 randomly generated controls. Furthermore, based on the hypothesis that physically interacting proteins share common biological functions, we extracted all the "one-hop" genes linking two autism risk genes. The significance of the extracted "one-hop" genes was further evaluated by analyzing their spatial and temporal expression patterns in the developing brain. 

Results:

The PIN dataset has grown rapidly since its first release in June 2011. As of September 2013, it includes 184 genes, 23214 interactions, and 1516 references. Our multi-faceted analysis of AutRef84 dataset containing rare and syndromic genes related to autism revealed small-world characteristics of ASD networks. Compared to 1000 controls of equal size (number of nodes and edges), the direct interactions network showed higher clustering coefficient (CC), lower average shortest path length (ASP) and lower diameter (D), indicating tight connectivity among the ASD risk genes.  Among the numerous "one-hop" proteins extracted from the AutRef84 main interactome, a significant number of them overlapped with existing ASD candidate genes.  Coexpression analysis using BrainSpan database indicated enriched connectivity among these candidates in brain region subsets at specific time periods, signifying their convergent role during development.

Conclusions:  PIN module serves as an invaluable resource for several bioinformatic analyses. Starting with a small list of seed genes, our analysis resulted in several convergent candidate autism genes which further showed spatial and temporal specificity aligned to the specific gene expression patterns observed during development. The PIN module and network analysis of similar nature would serve as a framework to guide further autism research and enable identification of diagnostic and therapeutic targets .

The PIN module integrated into AutDB  is licensed to Simons Foundation as SFARI Gene and is available at https://gene.sfari.org/autdb/PINHome.do .