An Alignment-free Regression Approach to Estimating Allele-Specific Expression in F 1 Animals

We wish to study allele-specific expression in diploid organisms, specifically in F1 animals with inbred parental strains. Current methods for analyzing allele-specific expression rely on read alignment, which leads to reference bias unless there is prior knowledge of all genomic variants in the parental strains. However, in the case where RNA-seq data is available for both parental strains, we do not need prior knowledge of parental genomic variants. Our approach first uses parental RNA-seq reads to create maternal and paternal versions of transcript sequences, then estimates allele-specific expression levels in the F1 animal for each transcript. Using the parental versions of all candidate transcripts as features, we use a modified lasso penalized linear regression model for estimating abundance levels of expressed transcripts in the F1 animal. We tested our methods on synthetic data from the mouse transcriptome and compared our results with those of Trinity, a state-of-the-art de novo RNA-seq assembler. Our methods achieved much higher sensitivity and specificity in both identifying expressed transcripts and transcripts exhibiting allele-specific expression. We were also able to separately predict relative expression levels from paternal and maternal strains with more accuracy.