Allow multiple trim lengths or read structure in TrimFastq

[bkohrn]

It would be nice if we could either give multiple output read lengths (one per file) or a read structure (similar to what exists in FastqToBam) in TrimFastq.

[nh13]

@bkohrn wouldn't be too hard to change this line to take in multiple lengths:

fgbio/src/main/scala/com/fulcrumgenomics/fastq/TrimFastq.scala

Line 49 in 5aded50

@arg(flag='l', doc="Length to trim reads to.") val length: Int,

, then propagate that further.

[eboyden]

+1 for read structure, so that 5' ends can be modified without having to run DemuxFastqs or FastqToBam

[nh13]

What's the objection to just using FastqToBam instead of TrimFastq?

[eboyden]

Only that it requires bam output, so if you have downstream steps that require fastq input (e.g. 3' quality or dovetail trimming, and certain aligners e.g. BWA), you need to perform the additional step to convert the ubam back to fastq, that's all. Probably lower on the triage list than some other things, especially if there's a faster/easier means to implement the op's request.

[bkohrn]

Mostly if the end goal is fastq files to use in further analysis (say, I have a 10X run that was sequenced as 150 x 10 x 10 x 150 bp reads, and I need to trim it down as fastq files to 26 x 10 x 10 x 90 bp reads). If I use FastqtoBam, I then need to do a second step to convert back to fastq files before I can proceed with analysis. Not that it isn't doable (I did it last time, using samtools fastq to convert back to fastq files), but it would probably save some time to be able to do the trimming in one step rather than two.

[nh13]

I'd just pipe into samtools fastq like you're doing to be honest. Of course there's no need to use output compression, so set fgbio --compression 0 FastqToBam. ... | samtools fastq ... which can speed it up.