mozack/abra2

Score threshold and content of the printed INFO

camillebenoist opened this issue · 0 comments

Hello,

the pipeline I am working on use in parrallele consensus (based on unique molecular index) or classical non consensus reads.
So for a sample, I have in the same time the bwa aligned reads, and the bwa conensus reads (duplicates reads coming from the same initial fragment are merge into a single consensus read, and the base quality is adjusted).

Using ABRA2, I was able to correctly adjust CIGAR for a 71 bp del event. However, ABRA2 doesn't find the same results depending of the input reads (classical or merge reads). More precisely, for a given read with the exact same alignment/CIGAR and sequence before ABRA2 realignment, only the deletion present in the consensus reads (with higher baseQ) are corrected.

Exemple for 2 identical reads, one coming from the basic process and the other from the consensus process :

1 - For basic reads (no difference before and after ABRA2)
before ABRA2 --------------------------------------
A00514:721:HHC55DRXY:1:2126:13856:21167 147 chr7 116411750 60 102M26S = 116411564 -288 AACACAGTCATTACAGTTTAAGATTGTCGTCGATTCTTGTGTGCTGTCTTATATGTAGTCCATAAAACCCATGAGTTCTGGGCACTGGGTCAAAGTCTCCTGCGCTACGATGCAAGAGTACACACTCC FFFFFFFF:FFFF:FF:FFFFF:FFFFFFFFFFFFF:FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF:FFFFFFFFFFFFFFFFFFFFFFFF NM:i:0 MD:Z:102 AS:i:102 XS:i:20 RG:Z:D721R130

after ABRA2--------------------------------------
A00514:721:HHC55DRXY:1:2126:13856:21167 147 chr7 116411750 60 102M26S = 116411564 -288 AACACAGTCATTACAGTTTAAGATTGTCGTCGATTCTTGTGTGCTGTCTTATATGTAGTCCATAAAACCCATGAGTTCTGGGCACTGGGTCAAAGTCTCCTGCGCTACGATGCAAGAGTACACACTCC FFFFFFFF:FFFF:FF:FFFFF:FFFFFFFFFFFFF:FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF:FFFFFFFFFFFFFFFFFFFFFFFF NM:i:0 MD:Z:102 AS:i:102 XS:i:20 RG:Z:D721R130

2 - For consensus reads (71pb DEL correctly aligned after ABRA2)
before ABRA2--------------------------------------
Illumina:608133 147 chr7 116411750 60 102M26S = 116411564 -288 AACACAGTCATTACAGTTTAAGATTGTCGTCGATTCTTGTGTGCTGTCTTATATGTAGTCCATAAAACCCATGAGTTCTGGGCACTGGGTCAAAGTCTCCTGCGCTACGATGCAAGAGTACACACTCC NNNNNNNNNNNNNNNNMNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN NM:i:0 MD:Z:102 AS:i:102 XS:i:20 RG:Z:D721R130

after ABRA2--------------------------------------
Illumina:608133 147 chr7 116411750 60 102M71D26M = 116411564 -385 AACACAGTCATTACAGTTTAAGATTGTCGTCGATTCTTGTGTGCTGTCTTATATGTAGTCCATAAAACCCATGAGTTCTGGGCACTGGGTCAAAGTCTCCTGCGCTACGATGCAAGAGTACACACTCC NNNNNNNNNNNNNNNNMNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN YA:Z:chr7:116411008:844M71D663M MD:Z:102 RG:Z:D721R130 NM:i:71YM:i:0 YO:Z:chr7:116411750:-:102M26S AS:i:102 XS:i:20 YX:i:20

The only difference I found was the base quality (baseQ is increase if the consensus is good between duplicate reads), but in the command I launched, I precise --mapq 0 --mbq 0 to be sure to not filter reads because of their poor quality.
Here is the complete command line I used :

"""
java -Xmx35g -Djava.io.tmpdir=$TMPDIR -jar ${ABRA2JAR}
--ref hg19_base.fa
--in ${INPUT_BAM}
--targets $BED
--out ${OUTPUT_BAM}
--mapq 0 --mbq 0
2> Abra2.log
"""

Finally I wanted to compare the INFO of the given segment between the 2 realignments but I don't find how to interpret the differences I saw :

for basic reads------------------------
INFO Fri Nov 05 20:32:32 UTC 2021 PROCESS_REGION_MSECS: chr7_116411546_116412048 63 0 13 0

for consensus reads------------------------
INFO Fri Nov 05 22:17:42 UTC 2021 PROCESS_REGION_MSECS: chr7_116411546_116412048 148 0 24 0

What is the meaning of the 4 last numbers? I didn't find the information...

If you could please explain me what could cause such differences and how to interprete the INFO, that would be great!
I can provide more information if needed!

Camille