Genomic and bioinformatics solutions to the search for genetic determinants of common, heritable disorders
Conrad Gilliam and Andrey Rzhetsky
This project developed new bioinformatics-based approaches to identifying disease genes and gene pathways, in particular those that harbor heritable determinants of common, debilitating disorders of the nervous system. Whole-genome surveys of copy-number variation identified roughly 400 potential genes contributing to autism. We prioritized the study of these genes using our tools for mining data from the literature and a network-based method for exploiting genetic linkage data. This analysis established the association of specific genetic alterations with autism, including a widespread microdeletion on chromosome 16.
We made considerable progress in elucidating genetic networks and candidate genes for autism, by combining whole genome experimental surveys of copy number variations (CNVs) with whole genome genetic linkage analysis and bioinformatic approaches. We collaborated in the whole-genome CNV analysis of two large data sets, and identified approximately 400 candidate autism genes (Sebat 2007, Christian 2008).
We prioritized these genes using two MAGNet-developed tools: PhenoGO, which detects genotype-phenotype co-occurrences in databases and journal articles (Sam 2009) and a network-based method for increasing the power of genetic linkage studies to infer genes responsible for complex disorders (Rzhetsky 2006a, Iossifov 2008). Networks were automatically inferred from the literature using the GeneWays algorithm (Rzhetsky 2004, Rzhetsky 2006b). This analysis established association of the microdeletion at genome position 16p11.2 (Kumar 2008) and the sex-specific interaction of ITGA3 and SLC6A4 (Suresh 2009) with autism, as well as additional candidate genes.
In another major effort, we developed a novel method to survey whole-genome linkage data for multi-gene patterns of inheritance. We used large-scale molecular interaction networks to limit the gene interaction search space sufficiently to survey multi-gene inheritance without sacrificing statistical significance (Iossifov et al 2008).
In an experimental collaboration at the University of Chicago, we used comparative genomic hybridization (CGH) with whole-genome tiling comprising some 19,000 bacterial artificial chromosome (BAC) clones. This method has lower resolution but higher accuracy than previous methods in detecting CNVs. We identified 63 CNVs that occurred only once in autism cases and not controls, encompassing a total of 285 potential autism candidate genes. Comparing with the previous study showed four CNVs in common for the two studies, encompassing 33 candidate genes. The chromosome-16 microdeletion appears in both studies, and we predict that it occurs in about 0.6 - 1.0% of all autism patients (Kumar, submitted).