DAWN: a framework to identify autism genes and subnetworks using gene expression and genetics

Area of Science:

• Genetics
• Neuroscience
• Bioinformatics

Background:

• De novo loss-of-function (dnLoF) mutations are more prevalent in autism spectrum disorder (ASD) cases compared to unaffected siblings.
• Identifying genes associated with ASD risk is crucial for understanding its genetic basis.
• Integrating diverse data types can enhance the identification of ASD-related genes.

Purpose of the Study:

• To develop a novel algorithm, DAWN, for accelerating the identification of autism spectrum disorder (ASD) genes.
• To model rare genetic variations and gene co-expression data from critical brain regions.

Main Methods:

• Developed the DAWN algorithm, a hidden Markov random field model.
• Integrated exome sequencing data (rare variations) with gene co-expression data from the mid-fetal neocortex.
• Combined gene interrelationships with gene-specific data, including expression and estimated risk effects.

Main Results:

• DAWN identified 127 potential ASD risk genes and associated subnetworks.
• Validation experiments confirmed DAWN's efficacy in predicting known and novel ASD genes.
• The algorithm successfully predicted known ASD genes not used in its initial training.

Conclusions:

• DAWN effectively predicts ASD genes and subnetworks by integrating genetic and gene expression data.
• Findings suggest neurite extension and neuronal arborization are implicated in ASD risk.
• DAWN can be applied to identify genes for other complex disorders and utilize emerging ASD data.