I have a set of 400 nucleotide sequences and want to cluster them on basis of similarity. For clustering, I am expecting a similarity <=45% among members of a cluster. Also, there will be a few sequences that do not show similarity to any other member. Is there any clustering approach that allow us to set a cut-off for relation (similarity) between members? and can keep the members with very low similarity to a "unclustered" set?

I have generated the percentage identity matrix (400 x 400) using clustal-omega, and using this matrix for clustering by "affinity-propagation" approach but not getting good results.

p.s. I have had used "cd-hit" and "uclust" already but they are not recommended for cases when expected sequence similarity is below 70%.

Link to my question on BioStar - https://www.biostars.org/p/147913/


  • $\begingroup$ This comment doesn't address your explicit questions but may provide an alternative to "affinity-propagation clustering." Permutation distribution clustering is an approach rooted in complexity analysis, information theory and (dis)similarity of time series data that is applicable to sequences of nucleotides. Full description with R code available here ... jstatsoft.org/article/view/v067i05/pdf $\endgroup$ – Mike Hunter Nov 27 '15 at 12:28

Hierarchical clusterings are commonly cut at a threshold level of similarity, such as 45%.

Furthermore, you can use DBSCAN, with epsilon set to 45%.

Plenty of more choices, if you keep on looking.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.