If I shred the Sanger reads into overlapping reads, wouldn't I be lending a statistical certainty to the new reads that doesn't exist? Ie, If a sanger base is incorrect, and that base is used in 50 'fake' reads, don't I grant certainty to an incorrect base where it shouldn't exist?
We're currently looking at different assemblers, and I'm in the process of learning about/testing ALLPATHS-LG. At the present we have 14 lanes of illumina data (for a mammalian-sized genome), but don't yet have the fragment (= superread) library. Could I still get a decent assembly by artificially creating the fragment library from a library of Sanger reads?
Thanks!
1 answer
This would probably work, as long as you have (1) high quality across the whole read (2) enough coverage. For (1) the super-read is constructed by overlapping alignment of 100bp reads from a 180bp fragment, so that the lower quality 3' ends of each read overlap to produce an overall higher quality 180bp read. For (2) I think you could be in trouble, as you won't get the 50x required from a Sanger library, unless you shred your Sanger reads into overlapping 180bp reads to the sufficient depth. In this case it would probably be fine.
Does ALLPATHS use quality scores at all? If not, if there is a base with low quality in your Sanger read, you could replace it with "N" to mark it for correction/PCR-check later. If so, perhaps scale all your Q scores in the Sanger reads by the factor appropriate for the shredding depth, and ALLPATHS will treat them more correctly.
Log in to answer this question.
Anybody have any ideas if this is a problem?