Split read extraction

[fanx]

My 454 data contains 20% reads with >1 hits in Bowtie 2. These reads were extracted by samtools. Are there any tools that can be use to make these multi-hit reads become single-hit reads?

I have examined almost all programs for chimeric read or SV detection, and it seems all of these programs have no option to do the job described above. I wonder if same tools, like samtool or bedtool, can do this job.

Many thanks!

[chadn737]

You could try mapping with more stringent settings, but really there isn't anything you can do. Reads mapping to multiple locations happens when you have multiple locations with same sequence.

[fanx]

I actually did to increase mapping stringency that reduced multi-hit reads from 30% to 20%. Question is I know most of these are chimeric reads. Anyway, thanks.

[chadn737]

How do you know they are chimeric, particularly if they are multi-hit reads? Chimeric reads shouldn't map back under stringent settings using just bowtie 2.

[fanx]

chadn737: Thanks.
Two ways: 1, I examined a small subset of these reads and found they are chimeric. 2, I know that my amplification method prior to 454 is highly prone to chimera formation.

My aim is to separate such a chimeric read into 2 single-hit reads assuming it is 2-hit reads. If there is no existing tools, I am thinking about the only way to do this job is overlapping trimming prior to Bowtie 2 aligning. There are some nice tools available to do this job. However, this method is not perfect since extact breakpoints are not known.

Another possibility is samtools/bedtools but I am not very familar to these tools.