Poster #80 - Katie Barfield

Adapting a Probabilistic Alignment Method for Multi-Mapping Rescue in Ribo-Seq Data

Katie Barfield, Clemson University Ishika Verma, The Pennsylvania State University Shaun Mahony, PhD, The Pennsylvania State University

Multi-mapping is a persistent challenge in short-read sequencing, particularly in ribosome profiling (Ribo-Seq). The repetitive nature and limited length of reads (20-38nt) often lead to ambiguous alignments. This reduces the number of usable reads available to inform downstream analysis and limits the biological insights that can be derived from Ribo-Seq datasets. While several bioinformatic tools have been developed to rescue multi-mapping reads, none are formatted for Ribo-Seq processing. Here, we adapt and test Allo, a probabilistic alignment software originally developed for other sequencing data, to rescue multi-mapping reads in Ribo-Seq workflows. We integrate Allo into a standard Ribo-Seq pipeline and assess its performance using artificially multi-mapping reads derived by trimming uniquely mapped STAR alignments to 20nt. Application of Allo increased the number of assigned Ribo-Seq reads by 312%. Validation of Allo mapping demonstrated 76.61% of reads were correctly assigned, with accuracy projected to rise to approximately 85% following training of Allo's neural network on Ribo-Seq data. By enabling the recovery of informative reads that would otherwise be discarded, this approach substantially enhances the interpretability of Ribo-seq datasets. The integration of Allo requires minimal additional computation cost, yet provides access to previously inaccessible biological insights. This pipeline improves the detection of novel translation events, regulatory elements, and coding regions, thereby broadening the range of biological discoveries possible through Ribo-Seq.