Nucleotide Match With Unix Tools

Question

[Michael Schubert]

Today one assignment in a course I'm doing was finding all (CCG)*4 repeats in the Q arm of human chromosome 11. Since I'm having a little too much time on my hands after having done it with EMBOSS's fuzznuc I wanted to try a bash-only version an came up with

cat 11q.fa | sed '1d' | tr -d '\n' | tr -d '\r' | egrep -io '(CCG){4}' | wc -l

fuzznuc came up with 18 matches, the bash version only with 11. Neither does take into account reverse complement matches by default.

I suppose fuzznuc is correct, but can anyone spot an error in the bash version?

(edit: it's called fuzznuc, not fuzzynuc)

[Michael]

give us the data and parameters to fuzzynuk and the output, then maybe if somebody has way too much time we'll figure something out ;)

Answer 1

[brentp]

I'm not familiar with fuzzynuc, but it probably finds overlapping matches whereas most linux tools will not find overlapping matches (so if you have CCGCCGCCGCCGCCG that's 5 triplets and you actually have 2 distinct (CCG){4}'s but the RE engine will only find the first.

You could sorta check this by seeing how many matches you find with '(CCG){5,}'as the regular expression--though that will similarly underestimate if there are 6+ CCG triplets together.

[Michael Schubert]

I checked it right now, that was indeed the case. Thanks!

[Michael]

sounds good! But then it's kind of fuzzy which result is correct

[Michael Schubert]

Yes it is. But it is good to know how the program behaves in case I need it again :-)

Answer 2

[Michael]

Well, I don't know fuzzynuc, so I dont have the slightest glimpse, though I dare to make a guess anyway. fuzzynuc, that rings a bell. Could it be that it does fuzzy matching, while grep does exact matching? To lazy to try to figure it out myself though, because I dont have your data.

[Michael Schubert]

fuzznuc also prints out a matchtable which does only show exact matches, not approximate ones.