Thread: BFAST published
View Single Post
Old 11-20-2009, 02:55 PM   #14
Nils Homer
nilshomer's Avatar
Location: Boston, MA, USA

Join Date: Nov 2008
Posts: 1,285

Originally Posted by HTS View Post
Hi Nils,

It is nice to read your paper, and I hope that my question is not completely out of the context (admittedly I haven't got time to dig into the details of your implementation yet). Basically I am thinking about applications other than resequencing, where I just want to find all local alignments passing a certain criterion (sure, you can put an upper-limit on the total number of matches for practical reasons, but I can use any of the tools out there to filter out heavy repeats first). An example is that my reference contains 10 cDNA sequences (different splice isoforms) belonging to the same gene with their own header, and for a given read, I want to find out how many transcripts this read is mapped to. I can do this straightforwardly with BLAT (by parsing the PSL file since I have gene/transcript information in the header) but the reference itself becomes ill-defined for all the global alignment tools out there. I am wondering if your tool can be easily adapted to handle this situation (while still making use of the base quality information as opposed to BLAT). If yes, probably I can finally stop using the old favorite BLAT as well. Please feel free to give me any suggestions/comments (also include all people who read this). Thanks a lot!

-- Leo
Absolutely! A similar example is suppose you have prior evidence of RNA splicing between two novel exons in which you only know the approximate boundaries (say within 100bp). You can create all possible joins of the two 100bp windows and then align your reads to the artificial reference to see which join is the most likely.
nilshomer is offline   Reply With Quote