Genotype prediction of 336,463 samples from public expression data.

Area of Science:

• Genomics
• Bioinformatics

Background:

• Vast amounts of RNA sequencing (RNA-seq) data exist, but often lack associated genotype information.
• Genotype data is crucial for various genetic analyses, including variant prioritization and expression quantitative trait loci (eQTL) studies.

Approach:

• Developed a statistical model using reference and alternative read counts from Recount3 RNA-seq experiments.
• The model predicts genotypes at autosomal biallelic loci within coding regions.

Key Points:

• Achieved high prediction accuracy: 99.5% overall, 99.6% major allele, and 90.4% minor allele accuracy.
• Model is robust to tissue and study variations with sufficient sequencing coverage.
• Applied the model to genotype all Recount3 samples, creating the largest genotype-informed RNA expression repository.

Conclusions:

• Predicted genotypes from RNA-seq data are sufficient for population structure analysis (e.g., Principal Component Analysis).
• This work significantly enhances the utility of public RNA-seq data for genetic research.