Filtering SAM before further analysis

[Rachelly]

Hi all,

I use BWA for mapping PE reads. I see that some reads are mapped when one end is on the forward strand while the other end is on the reverse strand.
If the insert size is in the range I gave as input, they are set as being in a proper pair (for example flag = 147 / 99).

What I think of doing is filtering out these reads, since it doesn't seem biologically possible that one end would be from one strand and the other end of the same read would be from the other strand.

I was wondering if I'm missing something here?
Does anyone do any filtering on a SAM file before using it for further analysis?

Thanks,
Rachelly.

[Rachelly]

Correction:
Before I thought that the flags are referring to the reads after the reverse-compliment process, but it seems they refer to the reads as they were sequenced.
So my question stands, only with a small correction - the reads that I think should be filtered out are those that map to the same strand. For example 129 or 177.

Thanks,
Rachelly.

[ulz_peter]

just to get it right: you want to filter out those pairs, where one read is mapping to the same strand as the mated pair, am I right?

If that's the case continue reading: wether of you want to keep those reads or not depends on what you're looking for: e.g. for whole genome / whole exome data that could mean (apart from being a sequencing artifact) that these reads span an inversion...

[Rachelly]

Right..
Thank, ulz_peter!