I have 8000 protein sequences that I want to cluster based on similarity (not identity) and select the longest representative sequence from each cluster. I checked several tools like HiFix, SiliX, ClusTR but could not find the optimal solution. I want to do clustering as CD-Hit does to reduce dataset but based on sequence similarity rather that sequence identity.
Kindly try RapidMiner and Weka 3.6. i had done similar task but it is for signaling/speach packets. if you will feel any problem then welcome to write again and i will help you.
regards
Muddsair
You could try our new clustering algorithm
Here is the paper describing it, if you are interest sent me a mail and I will sent you an executable.
http://www.biomedcentral.com/1471-2105/14/213/abstract
Regards
Lazaros
Have you ever tried "uclust" (http://drive5.com/usearch/manual/uclust_algo.html)? Based on the description, I think it is a fit for your problem.
You may use the affinity propagation algorithm that takes as input a matrix of similarities between the points (your proteins) and output the clusters and an exemplar from each cluster. Matlab and C implementations are available from http://www.psi.toronto.edu/index.php?q=affinity%20propagation
- Badges
- Science topic
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics