NextSeq variable length reads

[biocomputer]

I've looked at about a hundred RNA- and ChIP-seq samples from HiSeq machines (from our lab, GEO, or ENCODE) over the last few years and within each sample every read length is always the same (ie. they're all 50 or 100 or whatever). I just looked at some data from a sequencing facility that's using a NextSeq, it's RNA-seq single-end 151 bp, but FastQC shows the sequences range from 35-151 (vast majority are 151). I asked the sequencing facility and they said this is normal, that the shorter reads are due to a combination of some fragments being too small and from adapter trimming. Does this sound right, is it anything to be concerned about? Are they doing or not doing some filtering or read processing that's not been done to most other sequencing data? Can most modern aligners deal with variable read lengths (currently using Bowtie2 and STAR)?

I've found some related posts, they both mention 35 as the minimum length so I'm guessing that's not a coincidence:

https://www.biostars.org/p/66866/

http://seqanswers.com/forums/showthread.php?t=28531

[dpryan]

This is a setting during demultiplexing. The bcl2fastq package from Illumina has an option to trim adapters when it demultiplexes samples. I get the feeling that this option isn't often used (I don't use it for any data produced at our institute), but apparently the facility you used does use this option.

Anyway, they should have just written all of this out for you.

[Brian Bushnell]

Illumina's adapter-trimming algorithm is pretty bad; I suggest you not use it and instead do it manually. Also, for 151bp reads, I highly recommend trimming the last base since it's always crap. You can do this with BBDuk with the flag "ftm=5" which will always trim any base that does not equal 0 modulo 5 (in other words, it will trim reads that are 151 to 154 bp long to 150 bp).