"Seekmer, a fast and extensible RNA-seq transcript quantification algorithm"
Quantifying expression levels of splice isoforms using RNA-seq data is essential for transcript-aware studies. Recent advances in alignment-free algorithms brought opportunities in developing fast isoform quantifiers with little sacrifice in accuracy. In this study, we present a fast isoform quantifier, Seekmer. By combining both de Bruijn graph and edit distance heuristics, Seekmer confidently estimates the expression levels of isoforms, especially those that are expressed at low levels. The algorithm was benchmarked against and outperformed both traditional alignment-based and recent alignment-free methods. Isoform network analysis revealed that applying estimation from Seekmer gave more robust analysis than other methods on real biological datasets. Cancer type prediction based on Seekmer’s estimation also gave more accurate results on subtypes with differentially expressed isoforms from the same genes. The hybrid approach adopted by Seekmer enables more extensions to this light-weight RNA-seq analysis tool.