filter for sequences shared among replicates

[criggs]

Hi all,

I have small RNA sequence data for 5 treatment groups with 4 replicates each. For each replicate, I want to filter the read lists to only include those sequence found at least once in each replicate. In other words, if the sequence has a count of 0 in one replicate, it should be removed from every replicate in that treatment.

Does anyone know how I could do this filter? Maybe there is a way in R using DESeq or genefilter? Or any other way? The lists are too long (1-2 million reads) to perform this in excel.

Thanks,
Claire

[Brian Bushnell]

By duplicate sequence, do you mean a read with another exact duplicate of that read in the same dataset? If so, you can do that with Dedupe in multiple steps using different set operations. Just start with a set of nonduplicate reads from one from replicate, and then for each other set, find the set intersection with the current set. After you do intersection 3 times you will only be left with reads shared by all sets. Finding each intersection takes a 3 steps (union then subtraction then intersection).

To make sets with only one copy per sequence (necessary first step):
dedupe.sh in=readsA.fq out=a.fq
dedupe.sh in=readsB.fq out=b.fq
...etc, for each replicate.

Set union (elements in A or B):
dedupe.sh in=a.fq,b.fq out=a_or_b.fq

Set subtraction (elements in B but not A):
dedupe.sh in=a.fq,a_or_b.fq uniqueonly out=b_minus_a.fq

Set intersection (shared members of A and B)
dedupe.sh in=b.fq,b_minus_a.fq uniqueonly out=a_and_b.fq

Dedupe is in BBTools.