HISAT2 and different Illumina versions/Phred quality scores

[biocomputer]

I have some RNA-seq data and FastQC says some of the files have quality scores encoded with "Illumina 1.5" and some with "Sanger / Illumina 1.9". HISAT2 has the option to specify --phred33 or --phred64. I think Illumina 1.5 is phred64 and Illumina 1.9 is phred33, is this correct?

For consistency in running my analysis pipeline I'm considering converting the 1.5 files to 1.9 before alignment, would this be a good/bad/neutral idea and what is a good tool to do this?

[gringer]

I wouldn't recommend trusting FastQC unless you have some other independent verification of the results.

When was the sequencing done? If it has been carried out in the last five years, it's most likely to be phred33. Here's my favourite reference that shows the differences:

https://en.wikipedia.org/wiki/Fastq#Encoding

It might be that the quality scores are so good that it has been detected as phred64. Converting is a bad idea: q40 values will become q10, which will mess up alignment and error correction.

[biocomputer]

It's two sets of sequencing from a year or two apart all done several years ago but I don't know exactly when, I can try to find out. But the samples from the first (older) batch are all indicated to be Illumina 1.5 while the second (newer) batch are all 1.9.

[gringer]

Okay, that counts as an independent reason for the difference.

It would be best to map the batches separately using different phred options in HISAT2. There may be batch effects associated with the different Illumina runs, and the process is simpler if it's done that way; the HISAT2 output can be merged post-mapping if desired (with 'samtools merge' on sorted BAM files).

The SAM format specification is that quality scores output should be phred+33, so the mapping process will auto-convert the phred+64 scores to phred+33.