Are secondary alignments (SAM) produced by STAR aligner linked with MAPQ score?

Question

[Claudio]

Hi everyone,

I am currently working with paired-end RNA-Seq data and I have problems in understanding some details of the SAM content. Let us use for instance one BAM file available in ENCODE

EXPERIMENT: https://www.encodeproject.org/experiments/ENCSR000AEW/

DIRECT DOWNLOAD: https://www.encodeproject.org/files/ENCFF602BYA/@@download/ENCFF602BYA.bam

The BAM file has been produced using STAR. From the header it is possible to retrieve the executed command:

@CO user command line: STAR --genomeDir out --readFilesIn ENCFF001ROL.fastq.gz ENCFF001ROK.fastq.gz --readFilesCommand zcat --runThreadN 8 --genomeLoad NoSharedMemory --outFilterMultimapNmax 20 --alignSJoverhangMin 8 --alignSJDBoverhangMin 1 --outFilterMismatchNmax 999 --outFilterMismatchNoverReadLmax 0.04 --alignIntronMin 20 --alignIntronMax 1000000 --alignMatesGapMax 1000000 --outSAMheaderCommentFile COfile.txt --outSAMheaderHD @HD VN:1.4 SO:coordinate --outSAMunmapped Within --outFilterType BySJout --outSAMattributes NH HI AS NM MD --outSAMtype BAM SortedByCoordinate --quantMode TranscriptomeSAM --sjdbScore 1 --limitBAMsortRAM 60000000000

From the ENCFF602BYA.bam file I filtered all the reads having read name "DFDF8JF1:270:C19DCACXX:2:1216:1336:52997". It produces 6 entries.

DFDF8JF1:270:C19DCACXX:2:1216:1336:52997    99  chr11   83874510    1   45M87726N54M65533N2M    =   84148457    274048  CAGTGTAATTTCTTCAAATTCATATTCAATTTCTGTCCCATTGACATAAGGAATTGTGTCCAAAGTATCTGTGTTGACAATTATAGGAGCAGGACTGGCCT   @@@FFDDFHHDFHGHGGJGIJJCIICII@FFHGIEHHEGFFG<@F<<FHDHGFHDGBFGHG@FEG??EFEHFGBC=)@@77BGGI4.?#############   NH:i:3  HI:i:1  AS:i:195    NM:i:0  MD:Z:101
DFDF8JF1:270:C19DCACXX:2:1216:1336:52997    355 chr11   83874510    1   45M87726N54M186096N2M   =   84148457    274048  CAGTGTAATTTCTTCAAATTCATATTCAATTTCTGTCCCATTGACATAAGGAATTGTGTCCAAAGTATCTGTGTTGACAATTATAGGAGCAGGACTGGCCT   @@@FFDDFHHDFHGHGGJGIJJCIICII@FFHGIEHHEGFFG<@F<<FHDHGFHDGBFGHG@FEG??EFEHFGBC=)@@77BGGI4.?#############   NH:i:3  HI:i:2  AS:i:195    NM:i:0  MD:Z:101
DFDF8JF1:270:C19DCACXX:2:1216:1336:52997    355 chr11   83874510    1   45M87726N56M    =   84148457    274048  CAGTGTAATTTCTTCAAATTCATATTCAATTTCTGTCCCATTGACATAAGGAATTGTGTCCAAAGTATCTGTGTTGACAATTATAGGAGCAGGACTGGCCT   @@@FFDDFHHDFHGHGGJGIJJCIICII@FFHGIEHHEGFFG<@F<<FHDHGFHDGBFGHG@FEG??EFEHFGBC=)@@77BGGI4.?#############   NH:i:3  HI:i:3  AS:i:194    NM:i:0  MD:Z:101
DFDF8JF1:270:C19DCACXX:2:1216:1336:52997    147 chr11   84148457    1   101M    =   83874510    -274048 TGTCCACTAACCTGCGACCCAGCTTCTTAGCATACCAGATAGAGGCAAACATAGCGATGCCAATGGAAGGAATTCACAATGAAGCTGCCGCACAATCAACT   ###########BB;3(=BCC@>3>;):?7.)?>A;CAGIHDC=4.)=.)>A=AFGDD90*?4FBAGGGEC?:**9?9*H9EA2+2)E@8C:=ADD?DD??@   NH:i:3  HI:i:1  AS:i:195    NM:i:2  MD:Z:3G3G93
DFDF8JF1:270:C19DCACXX:2:1216:1336:52997    403 chr11   84148457    1   101M    =   83874510    -274048 TGTCCACTAACCTGCGACCCAGCTTCTTAGCATACCAGATAGAGGCAAACATAGCGATGCCAATGGAAGGAATTCACAATGAAGCTGCCGCACAATCAACT   ###########BB;3(=BCC@>3>;):?7.)?>A;CAGIHDC=4.)=.)>A=AFGDD90*?4FBAGGGEC?:**9?9*H9EA2+2)E@8C:=ADD?DD??@   NH:i:3  HI:i:2  AS:i:195    NM:i:2  MD:Z:3G3G93
DFDF8JF1:270:C19DCACXX:2:1216:1336:52997    403 chr11   84148457    1   101M    =   83874510    -274048 TGTCCACTAACCTGCGACCCAGCTTCTTAGCATACCAGATAGAGGCAAACATAGCGATGCCAATGGAAGGAATTCACAATGAAGCTGCCGCACAATCAACT   ###########BB;3(=BCC@>3>;):?7.)?>A;CAGIHDC=4.)=.)>A=AFGDD90*?4FBAGGGEC?:**9?9*H9EA2+2)E@8C:=ADD?DD??@   NH:i:3  HI:i:3  AS:i:194    NM:i:2  MD:Z:3G3G93

I understand most of the information reported here. Because it is paired-end data, each read name is present at least twice. Once per mate. From this example there is one primary alignment (first/fourth entry) and two secondary alignments (second/fifth entry, third/sixth entry). The multiple alignment is expressed by the different CIGAR score in the first three entries.

I am investigating where reads split across the genome, so to identify those exons that are expressed in the same isoform (de novo analysis). Now my questions are three:

1. Out of the three multiple alignments, should I take the primary one while considering the coverage?
2. Is the MAPQ score related with the number of multiple alignments? In the STAR documentation, "1" means "Maps to 4-9 locations in the target", but here I see only three multiple alignment. Does MAPQ involve also the CIGAR information?
3. "AS" SAM tag express the "Alignment score generated by aligner", how can I put the scores provided in this example in context?

Thanks in advance for the help

Claudio