Adding k-mer and q-mer counting. #269
Conversation
|
All automated tests passed. |
| val q = w.map(_._2) | ||
|
|
||
| // reduce bases into string, reduce quality scores | ||
| (b.map(_.toString).reduce(_ + _), q.reduce(_ * _)) |
There was a problem hiding this comment.
Is there an underflow / floating-point issue here? Is there a reason you're not just doing this computation in log-space? (i.e. omit the toSuccessProbability call above, and then q.reduce(_ + _) here instead?)
There was a problem hiding this comment.
There isn't an underflow issue when you're looking at short k-mers. If you have a 100-mer with all bases at Q20, your q weight for the q-mer is about 0.35.
Also, underflow isn't such an issue for q-mer counting. The goal is to get an expectation for the count of a k-mer, therefore q weights that underflow wouldn't have contributed much to the expectation anyways.
There was a problem hiding this comment.
Also, a 100-mer would be exceptionally long; current k-mer counting tools are pretty stoked about counting 20-mers.
There was a problem hiding this comment.
Fair enough -- okay, I'll merge as is.
|
See my question about log-scale computation, above (which might have some perf implications too, right?) If you address that and make sure this is rebased off of master, I'd be willing to merge this. |
|
@tdanford see my comments inline, I don't think there's a good reason to move to log scale. |
|
When you rebase, then, I will merge. |
| @Argument(required = true, metaVar = "KMER_LENGTH", usage = "Length of k-mers", index = 2) | ||
| var kmerLength: Int = 0 | ||
| @Args4jOption(required = false, name = "-countQmers", usage = "Counts q-mers instead of k-mers.") | ||
| var qmersNotKmers: Boolean = false |
There was a problem hiding this comment.
Naming this countQmers is more clear.
|
Okay, if you make the changes suggested by Matt and rebase off of master, then I will hit the 'merge' button :-D |
|
All automated tests passed. |
Adding k-mer and q-mer counting.
|
Thanks, @fnothaft! |
This adds functions for k-mer and q-mer counting, which is necessary for error correction. Additionally, this adds commands that support counting k/q-mers from the CLI.
q-mers are quality score weighted k-mers, and are described in http://genomebiology.com/2010/11/11/R116/abstract.