CrossFilt: A Cross-species Filtering Tool that Eliminates Alignment Bias in Comparative Genomics Studies

Abstract

Comparative functional genomic studies are often affected by biased read mapping across species due to inter-species differences in genome structure, sequence composition, and annotation quality. We developed CrossFilt, a filtering strategy that retains only sequencing reads that map reciprocally between genomes, ensuring that quantification of read counts is based on directly comparable genomic features. Using both real and simulated RNA-sequencing data from primates, we show that CrossFilt outperforms five alternative approaches that are commonly used, resulting in more accurate inference of gene expression differences. Our results highlight the impact of preprocessing strategies on the analysis of cross-species functional genomics data.