[Dock-fans] Are there some free programs to do the similarity clustering?

[John J. Irwin]

Hi Yolanda

Clustering can be helpful to reduce a complex dataset to patterns. But 
clustering is also in the eye of the beholder, with all the subjectivity 
that implies.

What is the question you want to answer? If it is "which compounds 
should I buy from among the top docking hits?", then in addition to 
clustering, you might want to think about the  underlying distribution 
of compounds in the database.

In our experience, perhaps surprisingly, we tend to cluster by eye. What 
I mean is that during a hit picking party, as we review the best looking 
compounds from among the top 500, if we see one that looks like one we 
saw before, we just say "NEXT!" and move on. If looks like one we said 
we would buy, then we consider this new compound as perhaps an 
alternative if the best scorer turns out to be unavailable, or expensive.

That's a bit of a non-answer, so here is how we do clustering: we use 
SUBSET1.0 from Marc Nicklaus (which is free). It is very fast, and, with 
cutoffs at 90, 80, and 70% Tanimoto, often gives us answers we can 
relate to.  Useful, yes, interesting, yes, but we still look at every 
one of the top scoring 500 before we buy.

Hope this helps.

John
UCSF DOCK Team

Äî Áõ wrote:
> Dear Dock-fans,
>  
> As a novice in virtual screening, I'm puzzled about how to do the 
> compound diversity analysis .
> Refering to some literatures, final results from DOCK should be 
> subjected to similarity clustering. Because still in attempt phases, I 
> expect that veterans in DOCK may recommend some utilities or free 
> programs to complete this experiment. Or is there some other methods 
> to solve this problem?
> Any help would be appreciated.
>  
> Best regards,
>
>
> Yolanda Guo
> Northeast Normal University  
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Dock-fans mailing list
> Dock-fans at docking.org
> http://blur.compbio.ucsf.edu/mailman/listinfo/dock-fans
>