Algorithms to Discover Structural Variation and Gene Networks in Neurodevelopmental Disorders

ABSTRACT

Algorithms to Discover Structural Variation and Gene Networks in Neurodevelopmental Disorders

The comparison of human genomes shows that along with single nucleotide variants and small indels, larger structural variants (SVs) are common. It is known that rare SVs can be a significant causal factor for complex disorders such as autism spectrum disorder (ASD), intellectual disability (ID), and cancer. I will first discuss methods to discover structural variation using high-throughput sequencing technologies and application of these methods to understand the role of rare SVs in disease. I will provide a combinatorial formulation for this problem under the maximum parsimony assumption and present an O(log n) approximation algorithm (VariationHunter). I will also present a combinatorial algorithm (CommonLAW), which allows simultaneous discovery of SVs in multiple genomes and, thus, improves the final comparative result among different individuals. We have successfully applied our methods to diverse sets of data including the Great Ape Genome Project to assess evolution, the 1000 Genomes Project to understand population variation, and a set of autism trios to characterize disease-relevant mutations. In order to assess the importance of rare SVs in autism, we performed deep genome sequencing (>50-fold sequence coverage) on 19 autism trios and used CommonLAW for SV discovery. Interestingly, we found rare and private SVs affecting autism-relevant genes such as SCN2A, CACNA2D4, and ARID1B.

Finally, despite extensive genetic heterogeneity underlying disorders such as ASD and ID, there is compelling evidence that risk genes are restricted to a small number of sets of interacting genes, or network modules. I will present a novel computational method (MAGI) that interprets de novo mutations in samples with autism using protein-protein interaction networks and brain-development RNAseq expression profiles to discover network modules relevant to neurodevelopmental disorders. Applying the method to recent de novo mutations from 1116 ASD and ID patients, we discovered two distinct and significant modules (p