Enter Note Done

Media

Decrease font size Increase font size

PgmNr 141: Spatiotemporal gene expression pattern predicts of autism risk genes.

Authors:
S. Chen 1,2; Y. Shen 1,2,3

View Session  Add to Schedule

Affiliations:
1) Integrated Program in Cellular, Molecular, and Biomedical Studies, Columbia University Irving Medical Center, New York, NY.; 2) Department of System Biology, Columbia University Irving Medical Center, New York, NY; 3) Department of Biomedical Informatics, Columbia University Irving Medical Center, New York, NY


Large scale exome sequencing studies have established that autism spectrum disorder (ASD) is a condition with strong but heterogenous genetic causes. Our knowledge of ASD risk genes is far from complete. We hypothesize that ASD risk genes have distinct spatiotemporal expression signatures in developing human brain under normal conditions.In this study, we obtained single-cell RNA-seq data of human fetal brain samples from a range of developmental stages in recent publications to infer brain cell-type specific gene expression. Using these data, we developed a new method, Frisk, to predict plausibility of ASD risk of all genes by Gradient Boosting. We used known ASD risk genes from SFARI Gene database as positives and the genes with de novolikely-gene disrupting (LGD) variants in unaffected siblings from the Simons Simplex Collection study as negatives.We assessed the performance by the ability to prioritization of de novo mutations in unknown risk genes, using data of 5964 cases from published ASD studies. Excluding all positive training genes, we selected gene sets based on ranking of Frisk score or other published methods. In each gene set, we calculated the enrichment rate of mutations in cases by comparing observed number with expectation from background rate, and in turn estimated precision and recall in the gene set. Frisk achieves higher enrichment, precision, and recall in prioritizing LGD and deleterious missense (Dmis) variants than other methods. Most of known ASD risk genes are intolerant of loss of function variants, as quantified by ExAC pLI≥0.9. In the genes with pLI<0.9, there is a trend of enrichment of LGD and Dmis variants (p=0.08). But in genes with pLI < 0.9 and Frisk score >0.4, such variants are significant enriched (p=4e-7), implicating ~60 candidate risk genes. These genes likely contribute to ASD risk through previous under-studied biological mechanisms or genetic models. Finally, we observed that high Frisk score is correlated with high expression in GABAergic and dopaminergic neurons from midbrain in late first trimester, GABAergic and excitatory neurons from prefrontal cortex in second trimester. With the unprecedent resolution of single-cell transcriptomics, our method will facilitate systematic discovery of novel risk genes and understanding of pathogenesis of ASDs.