Filtering of RNA fastq files prior to aligment

Question

[ognjen011]

I have a pair of RNA fastq files that are heavily contaminated with rRNA sequences, to such extent that almost one half of reads originate from a 10K long sequence of reference. I want to remove these reads, and I know I can do it at BAM level in one of many ways (ike Samtools View), but I am curious if there is an efficient way of removing those reads at the fastq stage. The idea is to speed up the alignment in some way.

Is that a good way to remove unwanted reads prior to alignment, in order to speed up the processing? Maybe a trimming tool, or such?

Answer 1

[Brian Bushnell]

The most efficient way would be alignment-free kmer-matching. To do that, you can remove all reads containing ribosomal kmers with BBDuk (where ref is the ribosomal sequence of this organism):

bbduk.sh in1=r1.fq in2=r2.fq out1=clean1.fq out2=clean2.fq ref=ribo.fa k=31 mm=f

You can alternatively gather a set of kmers that are present in ribosomal sequence but not in the rest of the genome and use that, but it's more complicated and probably not necessary.

[ognjen011]

Thank you! Although this is exactly what I need, is there a free to use alternative? BBDuk seems to be a strictly commercial software.

[ognjen011]

Sorry, I seem to be looking at the wrong package. Thanks for the answer!

[Brian Bushnell]

To make this clear to future readers, BBDuk is open-source, available here, and free for all uses.