T2G: fast and accurate transcriptome to genome BAM liftover for short and long read RNA sequencing

Abstract

Aligning RNA sequencing reads to the transcriptome rather than the genome recovers short and junction-spanning reads that genome alignment loses or misplaces, but it leaves the alignments in transcript coordinates, which genome browsers, variant callers and most annotation tools cannot use. Existing liftover tools translate between genome assemblies, and the few that reach transcript coordinates are tied to a single quantification model or assay type. Here we describe T2G, which converts complete BAM files from transcriptome to genome coordinates while keeping CIGAR strings, alignment tags and per-base RNA modification calls intact. Across nanopore direct RNA sequencing, ribosome footprinting and paired-end Illumina RNA sequencing, aligning to the transcriptome and lifting with T2G gave higher on-target specificity and more accurate splice junctions than direct genome alignment, removed the systematic assignment of reads to processed pseudogenes, and preserved isoform identity, revealing isoform-specific modification levels that genome alignment pools together. In a reanalysis of a published RNA-binding protein experiment, T2G reduced the number of reads attributed to pseudogenes 8.5-fold compared with unspliced genome alignment. T2G is available at https://github.com/Arnaroo/T2GD as a Python reference implementation and as t2gd, a dependency-free binary for Linux, macOS and Windows that also provides a gene-level quantifier and BAM utilities toolkit.