hi,
How can i do heirarchical clustering of protein sequences which are in fasta format. Is there any software available ?thanks in advance.
7 answers
-
if you want to cluster based on sequence identity then use CD-HIT
-
Based on phylogeny: Align Fasta sequences using clustal or any other Alignment software. then use any phylogenitic packaeges like MEGA or phylip etc. sequence-alignment-distance matrix (protdist)-then use NJ or parsimony or ML then generate phylogenetic tree and look for sequences falling in to different brances (they are clustered together)
-
Use the alignment and get the similarity or distance matrix for each pairs and use that as an input to any clustering softwares like Eisen's Cluster3 or Tm4 MEV. ( some tweeking is required to convert the data into appropriate format)
Use a spectral method instead?
This is a shameless self plug since I did this a few years back.
uClust was also released recently: http://drive5.com/usearch/usearch3.0.html and the publication
Exactly as Paulo writes - more info is needed. Do you want to cluster in a phylogenetic manner? By presence/absence of functional (Pfam) domains? By genetic effect(s) of knock-out/over-expression? Asking a specific question will get you a specific answer - and the help you need to move your research forward.
Eisen has some nice clustering prorgams: http://rana.lbl.gov/EisenSoftware.htm
I got very good results with MC-UPGMA: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0013409
It is also very efficient on large data sets.
You can download the program here: http://www.protonet.cs.huji.ac.il/mcupgma/
Log in to answer this question.
Hierarchial Clustering