Check out this BMC Bioinformatics paper entitled "Short clones or long clones? A simulation study on the use of paired reads in metagenomics"
"This paper addresses the problem of taxonomical analysis of paired reads. We describe a new feature of our metagenome analysis software MEGAN that allows one to process sequencing reads in pairs and makes assignments of such reads based on the combined bit scores of their matches to reference sequences. Using this new software in a simulation study, we investigate the use of Illumina paired-sequencing in taxonomical analysis and compare the performance of single reads, short clones and long clones. In addition, we also compare against simulated Roche-454 sequencing runs."
"Our study suggests that a higher percentage of Illumina paired reads than of Roche-454 single reads are correctly assigned to species."
"The gain of long-clone data (75 bp paired reads) over long single-read data (250 bp reads) is still significant at ≈ 4% (not shown)."
of course more importantly
"The authors declare that they have no competing interests."
I am not sure if such a program exists but I wonder if there is a aligner that takes into account the size between mate pairs and paired ends. Theoratically it should improve mapping. but by how much is unknown
Most good alignment tools take into account the mate distance (BFAST/BWA/MAQ/etc) and some even implement a "mate-rescue" (BWA). These not only improve the number of reads that map, but more importantly their confidence. The question I would ask is can you overfit in such situations such that you miss translocations and other structural variation?
ReplyDelete