Help understanding BBMAP output

Question

[m.radz]

I am mapping illumina PE total RNA-seq reads to a human reference with the aim to keep and analyse the unmapped reads for microbial content. This is the command I am using:

bbmap.sh in1=R1.fastq.gz in2=R2.gastq.gz outm=mapped.fastq.gz outu=unmapped.fastq.gz maxindel=200k ambig=random intronlen=20 xstag=us

The out I get from the mapping is this:

Read 1 data:            pct reads       num reads       pct bases          num bases

mapped:                   1.4446%         4616012         1.4214%          444831621
unambiguous:              0.5291%         1690797         0.5257%          164530054
ambiguous:                0.9155%         2925215         0.8957%          280301567
low-Q discards:           0.1385%          442659         0.0495%           15493065

perfect best site:        0.2236%          714613         0.2271%           71086566
semiperfect site:         0.2257%          721039         0.2292%           71728958
rescued:                  0.0340%          108781

Match Rate:                   NA               NA        45.0251%          437905504
Error Rate:              84.5007%         3900565        54.9729%          534655405
Sub Rate:                83.8672%         3871322         0.6738%            6553495
Del Rate:                 3.8987%          179964        54.2627%          527748664
Ins Rate:                 1.6776%           77440         0.0363%             353246
N Rate:                   0.2227%           10282         0.0020%              19376
Splice Rate:              2.6624%          122895       (splices at least 20 bp)


Read 2 data:            pct reads       num reads       pct bases          num bases

mapped:                   1.4562%         4653230         1.4351%          449664382
unambiguous:              0.5736%         1832980         0.5704%          178723715
ambiguous:                0.8826%         2820250         0.8647%          270940667
low-Q discards:           0.1382%          441542         0.0498%           15597422

perfect best site:        0.2657%          849081         0.2699%           84561207
semiperfect site:         0.2661%          850225         0.2702%           84672337
rescued:                  0.0534%          170740

Match Rate:                   NA               NA        49.6060%          442644534
Error Rate:              81.7453%         3803799        50.3914%          449653141
Sub Rate:                81.0799%         3772836         0.7479%            6674030
Del Rate:                 3.6818%          171322        49.6073%          442656394
Ins Rate:                 1.5347%           71415         0.0362%             322717
N Rate:                   0.1796%            8358         0.0026%              23101
Splice Rate:              2.5801%          120056       (splices at least 20 bp)

Total time:             35599.792 seconds.

There is a low mapping rate which I am happy with, I wanted minimal human contamination as possible. But I should I be concerned about the Match Rate for read 1 (NA) and the high Error Rate for both read 1 and read 2 (>80%). What exactly does the error rate mean?

Answer 1

[GenoMax]

There is no detailed explanation of how Brian calculates these myriad stats. This old post from Brian at SA is about the closest you are going to get: https://www.seqanswers.com/forum/bioinformatics/bioinformatics-aa/54070-error-rate-in-bbmap?p=278378#post278378

I personally only look at the mapped percentages of read 1 and read 2 along with the % of properly paired reads.

I am going to say that your results look good. You can take a few reads from unmapped pool and blast them at NCBI to see what you get. They should be non-human hits.

[m.radz]

Thanks Geno, On that this is the pairing data:

Pairing data:           pct pairs       num pairs       pct bases          num bases

mated pairs:              1.2423%         3969692         1.2196%          763829132
bad pairs:                0.0347%          110810         0.0338%           21194634
insert size avg:          288.05

Is the percent pairs a concern?

[GenoMax]

No because your number matches closely (will generally be a few points lower) with your alignment % for Read 1 and 2.

[m.radz]

Ahh ok, so the pairing data is that just for the reads that map to the reference?

[GenoMax]

Mated pair data indicates that the reads mapped to the reference at an expected distance.