🐛 markduplicates has issues with cram files

[bunop]

By visually inspecting the alignments (using samtools tview, there are some regions which seems to have bad alignments: the reason seems to be that Markduplicates changes the sequence in aligned file. For example, before markduplicates we have:

samtools view WT.cram Chr01:60000-60001 --reference Pvulgaris_442_v2.0.fa | head -n1
A01083:294:H3LHWDSXC:2:2553:27751:35556 147     Chr01   59859   60      150M    =       59547   -462    CGCCGCGTCTTTTAAGAAAATAGCGGGAGAAGAAACTTCGATTTTCAATAACAATGAAGGTAAATTAAATTGATAAATTTTATATTCAATTGATAGCAATAAATCACGCAAATATGTAAATTGAAATATTTATTTTAAAGTTTCGATAAC FFF:,FFF:F:FFF,FFFFFF:,F,,,:FFFFFFFFFFFFFFFFFFFFFFFFFFFFFF:FFFFFFF:FFFFFFFFFFFFFFFFFFF:FFF:FFFFFFFFFFFFFFFFFFFF:FF:FFFFFFFFFF,FFFFFFFFFFFFFFFFFFFFFFFFMC:Z:150M       AS:i:145        XS:i:20 MD:Z:24T125     NM:i:1  RG:Z:WT

And after markduplicates we have:

samtools view WT.md.cram Chr01:60000-60001 --reference Pvulgaris_442_v2.0.fa | head -n1
A01083:294:H3LHWDSXC:2:2553:27751:35556 147     Chr01   59859   60      150M    =       59547   -462    NCNNCNCGNGGGGCCCCCCCGCCNCCCCCCCCCCCNGGNCCGGGGNCCGCCNCCGCCCCCGCCCGGCCCGGCCGCCCGGGGCGCGGNCCGGCCGCCNCCGCCCGNCNCNCCCGCGCGCCCGGCCCCGCGGGCGGGGCCCCGGGNCCGCCN FFF:,FFF:F:FFF,FFFFFF:,F,,,:FFFFFFFFFFFFFFFFFFFFFFFFFFFFFF:FFFFFFF:FFFFFFFFFFFFFFFFFFF:FFF:FFFFFFFFFFFFFFFFFFFF:FF:FFFFFFFFFF,FFFFFFFFFFFFFFFFFFFFFFFFAS:i:145        MC:Z:150M       PG:Z:MarkDuplicates     XS:i:20 MD:Z:0C0G0C0C0G0C0G0T0C0T0T0T0T0A0A0G0A0A0A0A0T0A0G0C0T0G0G0A0G0A0A0G0A0A0A0C0T0T0C0G0A0T0T0T0T0C0A0A0T0A0A0C0A0A0T0G0A0A0G0G0T0A0A0A0T0T0A0A0A0T0T0G0A0T0A0A0A0T0T0T0T0A0T0A0T0T0C0A0A0T0T0G0A0T0A0G0C0A0A0T0A0A0A0T0C0A0C0G0C0A0A0A0T0A0T0G0T0A0A0A0T0T0G0A0A0A0T0A0T0T0T0A0T0T0T0T0A0A0A0G0T0T0T0C0G0A0T0A0A0C0    NM:i:150        RG:Z:WT

Then number of matches is identical, however markduplicates add 150 mismatches, and the sequence changed in column 10 is the sequence visualized using samtools tview. This behaviour does not affect all the genome regions. Is not clear how this affects the calling process. Markduplicates should be removed as described in #71

[bunop]

this bug seems to be fixed in picard release 3.2.0 which fix a bug related to CRAM files