DeepSAP: improved RNA-seq alignment by integrating transcriptome guidance with transformer-based splice junction scoring
摘要
Advancements in high-throughput sequencing have revolutionized transcriptomics, enabling insights into gene expression, splicing, and fusions. However, RNA-seq analysis remains challenging due to complex splice junctions, multi-mapped reads, and chimeric events. We present DeepSAP, which improves RNA-seq alignment by integrating GSNAP’s transcriptome-guided genomic alignment with transformer-based splice-junction scoring. This synergy enhances splice-junction detection, indel identification, and resolution of complex splicing patterns. On the Baruzzo human simulated benchmark across complexities, DeepSAP achieves the highest mean F1-score for splice junction detection, outperforming DRAGEN, novoSplice, STAR, HISAT2, and Subjunc. DeepSAP captures intricate sequence patterns surrounding splice donor and acceptor sites, advancing RNA-seq analysis.