CLASS: constrained transcript assembly of RNA-seq reads

RNA-seq has revolutionized our ability to survey the cellular transcriptome in great detail. However, while several approaches have been developed, the problem of assembling the short reads into full-length transcripts remains challenging.

We developed a novel algorithm and software tool, CLASS (Constraint-based Local Assembly and Selection of Splice variants), for accurately assembling splice variants using local read coverage patterns of RNA-seq reads, contiguity constraints from read pairs and spliced reads, and optionally information about gene structure extracted from cDNA sequence databases. The algorithmic underpinnings of CLASS are: i) a linear program to infer exons, ii) a compact splice graph representation of a gene and its splice variants, and iii) a transcript selection scheme that takes into account contiguity constraints and, where available, knowledge about gene structure.

In comparisons against leading transcript assembly programs, CLASS is more accurate on both simulated and real reads and produces results that are easier to interpret when applied to large scale real data, and therefore is a promising analysis tool for next generation sequencing data.

CLASS is available from http://sourceforge.net/projects/splicebox.