Clustering very short sequences

[graceqy]

Does anyone know any program that does clustering efficiently with super short sequences such as 3-4 amino acid long?

I was trying to cluster amino acid sequences by similarity and would like to have a similarity threshold 0.7. My sequences are 3-25 aa long. CDHit throws out sequences shorter than 6 and usearch is not accurate for short sequence cluster either. For example 'EFGV' and 'EFGH' by definition has 75% similarity, but with 0.7 similarity threshold usearch puts them into different clusters. I haven't been able to find another program that does what I want: to cluster short sequences by similarity.

Any thoughts?

[GenoMax]

How many sequences are there?

You could just give unix "sort" a try.

[graceqy]

I have tens of thousands of sequences. I need a similarity calculation that put sequences with > 70% similar to the same cluster.I don't think sort does that.

[GenoMax]

No it would not. Tens of thousands should not be a problem for sort and the results can give you an idea of how many clusters to expect. You may also want to "uniq" things out at the same time to remove redundancy.

You would need to code something to look at the sorted file and classify based on your criteria of > 70% (I assume for sequences of similar length).