bowtie2 : different MAPQ if -k is set

Question

[Nicolas Rosewick]

Hi,

I did some alignment of DNA libraries using bowtie2.

I tested with and without -k option.

Without -k options (-k 10 in this example), MAPQ ranged between 0 and 42 .

   MAPQ   Freq
1     0  29201
2     1 584134
3     2   6504
4     3   5377
5     4    251
6     5     27
7     6 144827
8     7  21386
9     8   6020
10   11   1481
11   12   2500
12   14    182
13   15    575
14   16     82
15   17    114
16   18    160
17   21     99
18   22    339
19   23   6940
20   24   7299
21   25     43
22   26     91
23   27     84
24   30   1305
25   31    266
26   32    269
27   34    456
28   35    790
29   36     22
30   37      8
31   38     41
32   39    341
33   40  13110
34   42 951540

With -k 10 : MAPQ ranged between 0 and 255

   MAPQ    Freq
1     0   40438
2     1 2023162
3     2   41425
4     3    3822
5     4     820
6     5     239
7     6  717882
8     7   29605
9    11   25711
10   12    1922
11   14    1548
12   15    4672
13   16     972
14   17     578
15   18     304
16   21     548
17   22    1554
18   25     257
19   26     661
20   27      26
21   30     818
22   31    1068
23   32    1327
24   34    1580
25   35    3890
26   36      78
27   37      35
28   38      80
29   39     350
30  255  430525

When I take the MAPQ 255 reads a look at their corresponding MAPQ in the "without -k" alignment I found MAPQ between 0 and 42 (most of them are 42). I do not understand why MAPQ will be differnet in -k mode ? I expected to have same MAPQ as -k mode will only output up to 10 (if -k 10) valid alignment per reads (stop me if I'm wrong ?).

   MAPQ   Freq
1     0   4053
2     1  29681
3     2    157
4     3    520
5     4      8
6     5      4
7     6   2420
8     7     73
9     8    773
10   11     72
11   12     14
12   14      4
13   15     27
14   16      4
15   21      1
16   22     12
17   23   1019
18   24   1142
19   26      4
20   30     72
21   31     17
22   32      1
23   34      2
24   39      2
25   40   3021
26   42 387953

FYI : bowtie2-2.2.9

Answer 1

[h.mon]

Although I don't recall exactly how mapping quality is calculated for Bowtie2 default mode (search BioStars, there were some threads about this in the recent past), the manual clearly states that the -k parameters alters quality mapping calculation:

-k <int>

By default, bowtie2 searches for distinct, valid alignments for each read. When it finds a valid alignment, it continues looking for alignments that are nearly as good or better. The best alignment found is reported (randomly selected from among best if tied). Information about the best alignments is used to estimate mapping quality and to set SAM optional fields, such as AS:i and XS:i.

When -k is specified, however, bowtie2 behaves differently. Instead, it searches for at most <int> distinct, valid alignments for each read. The search terminates when it can't find more distinct valid alignments, or when it finds <int>, whichever happens first. All alignments found are reported in descending order by alignment score. The alignment score for a paired-end alignment equals the sum of the alignment scores of the individual mates.