Removing PhiX reads - at what point?

[Harriet Hunt]

HI all,

no-one seems to have posted on this before - I discussed briefly with Julian last week, but maybe others have some experience...

I have a GbS library that was spiked with 30% PhiX and I'm wondering how and when to get rid of the PhiX reads in the analysis. Common seems to be to align to PhiX genome with Bowtie and keep unmatched reads. If I do this on the raw data, before any trimming, and then proceed to process_radtags, will adapter sequence and poor quality regions in the reads make it difficult to align reads to PhiX? Also, my data are paired-end reads... so I need to keep them in-phase for further analysis.

My species has a reference genome, so I guess I could leave the PhiX reads in there and they will get discarded by some combination of not having one of my barcodes and/or not aligning to my reference genome, but this will make the statistics on numbers of reads before/after trimming etc. a bit misleading.

If anyone can share how they've dealt with this issue I'd be very grateful...