Identifying discordantly mapped reads

Question

[Linda]

How can I use samtools to identify discordantly paired reads or is it even possible? I can use awk to simply output reads which are mapped on different chromosomes, and also reads where the distance between the pairs is significantly larger than expected but is there a way to do this via samtools?

[Pierre Lindenbaum]

just filter out flags 3854 ( http://broadinstitute.github.io/picard/explain-flags.html ) ? (

read mapped in proper pair
read unmapped
mate unmapped
not primary alignment
read fails platform/vendor quality checks
read is PCR or optical duplicate
supplementary alignment

)

Answer 1

[iraun]

There are different options:

- You can use samtools and give as argument the following flag '-F 1294'. In this way you'll discard all reads having which are: read mapped in proper pair, read unmapped,mate unmapped,not primary alignment,read is PCR or optical duplicate.
- As you've mentioned, you can use awk to extract mates mapped on different chr. You should convert your bam to sam, and then awk '($3!=$7 && $7!="=")' .

I would vote for the first approach, because it includes the distance between pairs issue.

Answer 2

[Pierre Lindenbaum]

using samjdk: http://lindenb.github.io/jvarkit/SamJdk.html

$ java -jar dist/samjdk.jar -e 'return record.getReadPairedFlag() && !record.getReadUnmappedFlag() && !record.getMateUnmappedFlag() && !record.getReferenceName().equals(record.getMateReferenceName());'  in.bam