ReadsMap: a new tool for high precision mapping of DNAseq and RNAseq read sequences

There are currently plenty of programs available for mapping short sequences (reads) to a genome. Most of them, however, including such popular and actively developed programs as Bowtie, BWA, TopHat and many others, are based on Burrows-Wheeler Transform (BWT) algorithm. This approach is very effective for mapping high-homology reads, but runs into problems when mapping reads with high level of errors or SNP. Also it has problems with mapping RNASeq spliced reads (such as reads that aligning with gaps corresponding intron sequences), the kind that is essential for finding introns and alternative splicing gene isoforms. Meanwhile, finding intron positions is the most important task for determining the gene structure, and especially alternatively spliced variants of genes. In this paper, we propose a new algorithm that involves hashing reference genome. ReadsMap program, implementing such algorithm, demonstrate very high-accuracy mapping of large number of short reads to one or more genomic contigs. It is achieved mostly by better alignment of very short parts of reads separated by long introns with accounting information from mapping other reads containing the same intron inserted between bigger blocks. Availability and implementation: ReadsMap is implemented in C. It is incorporated in Fgenesh++ gene identification pipeline and is freely available to academic users at Softberry web server www.softberry.com.