TransComb: genome-guided transcriptome assembly via combing junctions in splicing graphs

Transcriptome assemblers aim to reconstruct full-length transcripts from RNA-seq data. We present TransComb, a genome-guided assembler developed based on a junction graph, weighted by a bin-packing strategy and paired-end information. A newly designed extension method based on weighted junction graphs can accurately extract paths representing expressed transcripts, whether they have low or high expression levels. Tested on both simulated and real datasets, TransComb demonstrates significant improvements in both recall and precision over leading assemblers, including StringTie, Cufflinks, Bayesembler, and Traph. In addition, it runs much faster and requires less memory on average. TransComb is available at http://sourceforge.net/projects/transcriptomeassembly/files/.