RepeatMasker and BLAST

[andreanna05]

Hi all,

I am trying select sequences for capture enrichment using an unpublished genome of the same species available from NCBI as a reference. The reference genome seems to be pretty good quality (approximately right size, large scaffolds, etc).

I went through and picked my sequences of interest (~500kb), broke them up into ~100bp sequences, and then used BLASTn to see if there were any sequences that had multiple hits in the genome, and threw them out. I also BLASTed my selected sequences against themselves and threw out any that had multiple hits. Finally, I thought I would check these for low complexity regions, so I ran them through RepeatMasker online, selecting the most closely related species in the database. About 40% of my sequences were masked (~15% are LINE1).

RepeatMasker is obviously much more sensitive than BLAST, and repeats could potentially be more divergent than could be detected with BLASTn. But I would have thought I would have had more BLAST hits of my sequences to themselves or to the reference genome with such a large proportion being repetitive. Any thoughts about why this might be?

[Brian Bushnell]

I'm not entirely sure what your question is, but my brief experience with RepeatMasker is that it masked WAY more than was necessary. And it was incredibly slow so I didn't want to have to re-run it repeatedly to get the desired result. As a result I had to write my own masking program, which can mask a genome for low-complexity regions and also regions covered by a sam file. To generate a sam file that will mask repeats, you can, for example, shred the genome into 100bp fragments (as you did), align them against the unshredded genome requiring high identity, then filter the sam file to only contain multi-mapping reads. Then it can be used as input to BBMask.

I don't know how RepeatMasker works, but I would investigate further before deciding to mask 40% of the genome. If BLAST can't even detect the homologies, why would you want to mask them as repeats? BLAST is very sensitive. Though of course it depends on what kind of identity threshold would start to interfere with your enrichment.