[Bioperl-l] Loading Blast Report in a minimal way

[zhihuali]

Hi netters,

I'm using SearchIO to parse my blast reports. They are extremely huge, and not surprisingly, it's extremely slow and sometimes the system crashed due to memmory problem. As I can handle small reports quickly, it seems like a problem related to the way SearchIO works: it slurps the whole report into the memory and builds millions of objects.

I've checked old posts and some people used FastHitEventBuilder to build hit objects without any hsp objects. And some people suggested using tabular output of blast. But in my case I need to go to each of the hsps of each hit, parse the alignment, and gather the information needed if that hsp fits certain criteria, and then move on to the next hsp/or jump over to the next hit/ or exit the processing, according to the information I have already got. An ideal way would be to read one hsp at a time from the report to the memory.  Is there some way to modify SearchIO (or build another Search Event) to do this?

Thanks a lot!


Zhihua Li
_________________________________________________________________
用 Live Search 搜尽天下资讯！
http://www.live.com/?searchOnly=true