Covariance-based sample selection for heterogeneous data: Applications to gene expression and autism risk gene...

Area of Science:

• Genetics
• Neuroscience
• Bioinformatics

Background:

• Autism risk is linked to genetic mutations, with gene expression patterns offering clues.
• "Guilt by association" methods like DAWN leverage correlated gene expressions to identify autism risk genes.
• Previous analyses of the BrainSpan dataset faced challenges due to gene expression heterogeneity across brain regions and developmental stages.

Purpose of the Study:

• To develop a novel method, COBS (Covariance-Based sample Selection), for identifying larger, more homogeneous sample subsets.
• To enhance the power of autism risk gene detection by addressing sample heterogeneity.
• To improve the accuracy of gene-based risk prediction in autism.

Main Methods:

• Developed COBS to select samples with a shared population covariance matrix.
• Applied COBS to the BrainSpan dataset for downstream DAWN analysis.
• Utilized genetic risk scores from sequential data freezes (2014 and 2020) to validate COBS.

Main Results:

• COBS successfully identified larger and more homogeneous sample subsets.
• The COBS method improved the predictive ability of DAWN.
• Demonstrated enhanced detection of autism risk genes using older genetic risk score data with the newer data freeze.

Conclusions:

• COBS is an effective method for improving autism risk gene detection.
• Addressing sample heterogeneity through COBS increases the power of gene expression-based association studies.
• This approach enhances the utility of existing datasets for discovering genetic factors in autism.