Bowtie2 mapping algorithm about multiseed process

Question

[--panda--]

I know bowtie2 separate the 'reads' into several 'seed's before mapping as shown below. My question is that do they map all the 'seed's and then combine the results of all the 'seed's to find the best position for the 'read'? If not, how?

The mannual of bowtie2 to do not explain it clearly and their paper associated neither.

Read: TAGCTACGCTCTACGCTATCATGCATAAAC

Seed 1 fw: TAGCTACGCT(pos:1-10)

Seed 1 rc: AGCGTAGCTA

Seed 2 fw: CGCTCTACGC(pos:7-16)

Seed 2 rc: GCGTAGAGCG

Seed 3 fw: ACGCTATCAT(pos:13-22)

Seed 3 rc: ATGATAGCGT

Seed 4 fw: TCATGCATAA(pos:19-28)

Seed 4 rc: TTATGCATGA

Answer 1

[Devon Ryan]

Have a look at figure S1 from the bowtie2 paper. In short, the seeds are prioritized and extended (if possible) into full alignments. Those are then ranked to find a best alignment, which is reported with a MAPQ given according to the other alignments.

[--panda--]

Thanks! That's helpful