How to find out if mapping is correct/not

[mitochy]

Hi all.
I'm really new to sequencing.
I'm trying to map single end RNA-seq reads (short) to reference genome using (any) short read aligner. My question is, how do I find out if the program is mapping correctly using my data?
- input is some million reads of Ecoli K12 illumina RNA-seq data, 35bp in length
- ref is Ecoli K12 ~4.8MBp
- e.g. I use BWA 0.50 with bwtsw and it generated a sam file with position of each reads.

How do I know how correct the result in the SAM file is (statistically)? Or is there no way to find out and we have to trust result of correctness/ROC based on mapping known/simulated data (like this one?)

[dpryan]

Have a look at the MAPQ field (5th column) of the SAM or BAM output file. I don't know how bwa calculates that (particularly of transcripts rather than simpler DNA reads), so you may want to browse the source code if you really need to know the nitty-gritty.

[mitochy]

That is true. The thing is the 5th column is gotten from bwa's own calculation, and maq/bowtie might have different result from their own calculation. So I guess we can only trust ROC comparison based on simulation data.

[dpryan]

Pretty much. I know from experience that two different aligners can align the same sequence to the same location and give very different MAPQ scores. Since many aligners are open source, you could just look at how they generate the score and see if it satisfies your requirements for the statistic (though I doubt it will). But, yeah, the simulation data is probably what you'll need to go by unless you feel up to modifying an aligner.