All question mark quality scores on several studies

Question

[Jonathan]

I've stumbled upon several shotgun studies where all sample bases are ? (so 63 in phred+33). The first time I thought they might have been tampered with, but having just downloaded samples from 7 studies and 5 of them end up like this makes me wonder.

I thought it might be a certain type of quality binning, but I can't seem to find any binning that goes that high. Here are some example studies where this happened (I haven't checked all sequences, but 100% of the sequences I've checked in those studies are all ???).

SRP403740
SRP423365
SRP424298
SRP425931
SRP434700
SRP410115

I see no apparent pattern; different sequencer models (Hi seq and Novaseq), and all these studies seem independent.

Does anyone have an explanation for this?

[MatthewP]

Maybe they applied Q30 filter before uploading data?

[Jonathan]

Even so, is it even possible for the sequencer to have an error rate of less than 1 in a million for literally 100% percent of the bases?

[MatthewP]

I mean, maybe they drop those reads with < Q30 quality.

[Jonathan]

In this case that would have been a <Q63 filter; that doesn't really make sense, no? Is it really possible get 30M reads with quality >63 out of a MiSeq?

And again, it's 100% of bases with the exact same quality score, and this has happened over several independent studies.

[MatthewP]

I mean, fastq format adds 33 to present quality, filter tools will also add 33 to the cutoff before filtering.

[Jonathan]

Here is an example:

@SRR22424561.1/2
GTTCATCTTCTCGATTCGGCGGTGGATGTAGACGCAGTTGGCATCGAGTTTCCTGATGGTTTCGCCTTCAAGCGATACGCGGAAACGTATTACTCTGAGGGTGTATGTGTGTTGTGGTCCGATATTTACTATACATTCACGGCTATCAAA
+
??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
@SRR22424561.2/2
CGTTAAGCACATGCTGTGTGTACATGGCTTTTTTAATGATAACATAGTGTGGCGACAATGTTGTGAACCCCGAAAGTCCGGAAACGGACGGAGACGATATGATTCCTTGTTCGGTCAACCGGGTTGATGATGGCATCTATATGGAAAG
+
????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
@SRR22424561.3/2
ACCGCCTGAGCGGTAGCGTAGAGTATTTCTCGCGGAAGACCTCGGACATGCTCTAGTACAAACCGGTATCTCCTTGGCTGGGTTACTCGCGGCAGCCTATCAACGTGGGTTAGATGCGCAATGCCGGTGTCGAGATTCAGGCGTCCGCC

[GenoMax]

I looked at the data from the example you posted and I see normal scores:

$ more SRR22424561_1.fastq 
@1
AATCCAAAGATAACCGCGAAACTCCGGAAATCCCTTCCGTCAGTGACATCTGGAAAGCGGCGGACTTCAACGAACGTGAAGTATTCGACTATTATGGAATTGTATTCTTCGGACATCGCGACATGAGGCGCCTTTATCTTCGTAACG
+1
-AAFF7A-AJJJJ-FJJJJ7-FJJFJFJ7FFJJJJJ7F<F7<JJ--FF-FFFFJ<<77<---J<AJJFJJJJJFJFAAAA-FAJJ<-7-<F<FAJJJ-AJAAAFJJJ<7-F<JJJFF-7--7-7AFF7F---77FA<7-AFAA7<7A 

How did you retrieve the data?