Ability to exclude reads

[deprekate]

It would be useful to have a function to exclude sequences from a file. Basically it would be the exact opposite of subseq.

seqtk remove in.fq name.lst > out.fq

Only sequences with name NOT in name.lst are extracted.

[tseemann]

This can be done in Unix like this:

cat in.fq | paste - - - - | grep -v -f -F name.lst | tr "\t" "\n" > out.fq

See my blog post for more details: http://thegenomefactory.blogspot.com.au/2012/05/cool-use-of-unix-paste-with-ngs.html

[deprekate]

That is a cool trick. I always seem to find new ways to use paste. But as a fix I added a fork, we will see if it gets accepted. It would be valueable, as the highest rated method is to use a QIIME script, which requires installing the whole QIIME package. I just like the ability to do the same/opposite task with the same tool:

seqtk subseq in.fa good.list
seqtk exclude in.fa bad.list

I chose exclude over remove since remove is ambiguous, technically speaking subseq is removing (it removes the reads in name.lst from the in.fa). Which is why I would switch the subseq flag to extract, as the term is more apt (the term subseq is more descriptive of splicing a read). But that changes the behavior of the original code, and most developer do not like changing already established functionality.

[tseemann]

The traditional way in Unix to do these two types of subsets is to have a single command, but have a -v flag to "invert" the behaviour:

seqtk subset in.fq good.list > good.fq
seqtk subset in.fq bad.list > bad.fq

And the inverse:

seqtk subset -v in.fq good.list > bad.fq
seqtk subset -v in.fq bad.list > good.fq

It would be good if samtools faidx had a -v option.

[conchoecia]

For fasta: I think a legal fasta is char wrapped, so this is safer until seqtk has a -v option.

bioawk -cfastx '{printf(">%s\t%s\n", $name, $seq)}' in.fa | \
  grep -v -f name.lst | tr "\t" "\n" > out.fa

[edwwlui]

This can be done in Unix like this:

cat in.fq | paste - - - - | grep -v -f -F name.lst | tr "\t" "\n" > out.fq

See my blog post for more details: http://thegenomefactory.blogspot.com.au/2012/05/cool-use-of-unix-paste-with-ngs.html

Great solution! Btw I am suggesting an edit to be cat in.fq | paste - - - - | grep -v -F -f name.lst | tr "\t" "\n" > out.fq since grep -f -F might throw an error that -F is treated as a file after -f.

[y9c]

There is a -M (mask) argument in seqtk seq command, but the reads are soft masked, not being removed from the file.