SA tag in SAM file for chimeric reads

[jazz]

Hi,
I am using BWA-mem for split read alignment for my single end genomic DNA-seq from Illumina. I know that BWA uses SA tag for marking chimeric reads. When I manually BLAST individual reads with the SA tag I can clearly verify that they are indeed chimeras. However, I could not find details about the SA tag itself. What information is encoded in the SA field? I am posting an example of a chimeric read that maps to two separate genomic locations within the same contig (scf7180000067989)


HWI-ST387:139:C03WJABXX:5:2108:15315:193815 16 scf7180000067989 85156 60 60M41S * 0 0 TTGAAGTCAAGAAAGTGGTAAAGAGAGATTAATAGGGGTATCTCAGCTACAACAAATATTATATTAAATTAAATGGTTAATCTTGCTTTGCTCACCATAAA * NM:i:2 MD:Z:31G1C26 AS:i:50 XS:i:0 SA:Z:scf7180000067989,85273,-,54S47M,60,1;

HWI-ST387:139:C03WJABXX:5:2108:15315:193815 272 scf7180000067989 85273 60 54H47M * 0 0 AATATTATATTAAATTAAATGGTTAATCTTGCTTTGCTCACCATAAA * NM:i:1 MD:Z:11T35 AS:i:42 XS:i:22 SA:Z:scf7180000067989,85156,-,60M41S,60,2;

I am expecting a lot of genome rearrangements in the sample, so ultimately I want to isolate these reads that map to variant locations and identify the regions of microhomology, which could help identify the breakpoint. I am new to Bioinformatics so any inputs would be great.

Thanks in advance!

[dcameron]

It is defined in the SAM tags specifications document which was split out from the main SAM specs at the end of last year.

Note that bwa can write split alignments in which chimeric segments overlap (eg: alignments of 60M40S, and 50S50M for the same read)

>I am expecting a lot of genome rearrangements in the sample, so ultimately I want to isolate these reads that map to variant locations and identify the regions of microhomology, which could help identify the breakpoint. I am new to Bioinformatics so any inputs would be great.

I would recommend using one of the many (50+ at my last count) structural variant callers available that are designed to identify breakpoints. In terms of bwa SA tags, LUMPY is a caller that combines the bwa split read aligments with read pair and coverage information for breakpoint calling. If you're not happy to let the caller perform it's own split read analysis, then I would recommend GRIDSS (disclaimer: my caller) due to the lower false discovery rater compared to other callers. Other callers performing decently in my benchmarking results are manta, CREST and DELLY (with Pindel quite good if your focus is sensitivity).