CombiUgi 500k docked (finally!!)

1 view
Skip to first unread message

Rajarshi

unread,
Jul 13, 2007, 11:02:55 AM7/13/07
to UsefulChem
Hi, apologies for the previous false starts. I have finally got a
successful set of docking runs. The top 10 SMILES are valid and have
proper conformations.

The data and summary is available at http://rguha.ath.cx/~rguha/cicc/ugidock/

Jean-Claude Bradley

unread,
Jul 13, 2007, 11:13:50 AM7/13/07
to usefu...@googlegroups.com
Rajarshi,
No problem - research is messy. That is only bad if we stop correcting ourselves :)
Your compounds 5 and 6 are duplicates.
We'll get started with ordering what we need.

 

Rajarshi

unread,
Jul 13, 2007, 11:40:10 AM7/13/07
to UsefulChem

On Jul 13, 11:13 am, "Jean-Claude Bradley"


<jeanclaude.brad...@gmail.com> wrote:
>
> Your compounds 5 and 6 are duplicates.

Oops! Fixed now.

jcb

unread,
Jul 16, 2007, 6:10:06 PM7/16/07
to UsefulChem
Rajarshi,
How exactly are you determining the order of the top compounds? I
don't see a pattern from the numbers in any of the columns.

Jean-Claude

Rajarshi

unread,
Jul 16, 2007, 8:48:10 PM7/16/07
to UsefulChem

On Jul 16, 6:10 pm, jcb <jeanclaude.brad...@gmail.com> wrote:
> Rajarshi,
> How exactly are you determining the order of the top compounds? I
> don't see a pattern from the numbers in any of the columns.

I'm using the consensus scoring scheme that Openeye uses.

So basically each molecule is scored by 4 scoring functions
(Shapegauss, Chemgauss3 etc).

So I first rank all the molecules by Shapegauss and save the ranks.
Next I rank the molecules by Chemgauss3 and save the ranks. Similarly
for the other scoring functions.

Then the consensus score for a given molecule is the sum of the ranks
for the molecule across the 4 scoring functions.

As a result, molecules with smaller consensus scores are 'overall'
better than molecules with higher scores.

Now, this is just one way to do the ranking. If prior information
indicates that a certain scoring function is better suited to the
problem, one could simply use that scoring function. Since I have no
prior information a consensus ranking seems to be the way to go.

But I'm open to other suggestions for ranking (all the raw data is
available on the website)

Jean-Claude Bradley

unread,
Jul 17, 2007, 10:05:28 AM7/17/07
to usefu...@googlegroups.com
Got it - the consRank is the one we're using now to prioritize.
It would be interesting to get feedback from other docking experts down the road about what they think

 

Egon Willighagen

unread,
Jul 17, 2007, 11:09:24 AM7/17/07
to usefu...@googlegroups.com
On 7/17/07, Jean-Claude Bradley <jeanclaud...@gmail.com> wrote:
> Got it - the consRank is the one we're using now to prioritize.
> It would be interesting to get feedback from other docking experts down the
> road about what they think

No hands on experience with docking, but seen from results on
conferences, the error is rather high, which is why consensus ranking
works.

From a project point of view, however, I would suggest to not pick the
ten best ranked candidates, but to pick the 10 most structurally
diverse structures from the top 25 (or 50). Or, diverse in terms of
other properties. That should get you to a lead-like compounds, which
allows you to tune later. DISCLAIMER: I haven't explored the full
SMILES dataset, so maybe they are all structural so similar, that
doing a diversity analysis makes no sense :)

Egon

Rajarshi

unread,
Jul 18, 2007, 11:07:15 AM7/18/07
to UsefulChem

On Jul 17, 11:09 am, "Egon Willighagen" <egon.willigha...@gmail.com>
wrote:


> On 7/17/07, Jean-Claude Bradley <jeanclaude.brad...@gmail.com> wrote:
>
> > Got it - the consRank is the one we're using now to prioritize.
> > It would be interesting to get feedback from other docking experts down the
> > road about what they think
>

> From a project point of view, however, I would suggest to not pick the
> ten best ranked candidates, but to pick the 10 most structurally
> diverse structures from the top 25 (or 50). Or, diverse in terms of
> other properties. That should get you to a lead-like compounds, which
> allows you to tune later.

This is certainly a good idea. Also note that before doing the
dokcing, no drug-like or lead-like filters were applied. Though a
docking run will probably exclude the more obvious non-drug-like
compounds (in terms of size), it'd be something to keep in mind.

As for getting lead like compounds I'm not sure why such a diversity
based selection would provide us with lead-like compounds unless we
can 'inject' a lead-like example compound into the chemical space of
the say the top 100 compounds and see where the they lie with respect
to the 'injected' compound.

Jean-Claude Bradley

unread,
Jul 18, 2007, 5:45:55 PM7/18/07
to usefu...@googlegroups.com
Egon,
Here is a little summary of where we stand
 
We're going to make the top compounds that lend themselves to being made easily and cheaply.  Hopefully there will be enough diversity there.
 
Jean-Claude

 
--
Jean-Claude Bradley, Ph. D.
E-Learning Coordinator for the College of Arts and Sciences
Associate Professor of Chemistry

Egon Willighagen

unread,
Jul 19, 2007, 12:08:50 PM7/19/07
to usefu...@googlegroups.com
On 7/18/07, Jean-Claude Bradley <jeanclaud...@gmail.com> wrote:
> Here is a little summary of where we stand
> http://usefulchem.blogspot.com/2007/07/combiugi-time-for-synthesis.html

You're quite right. I should think before speak :) I did read that
item, but did not think about it when writing up my comment.

> We're going to make the top compounds that lend themselves to being made
> easily and cheaply. Hopefully there will be enough diversity there.

Still, I would guess that easily, cheap starting materials might still
give you enough flexibility...

Egon

Jean-Claude Bradley

unread,
Jul 19, 2007, 12:21:41 PM7/19/07
to usefu...@googlegroups.com
The order for the starting materials is going out today so we should find out soon :)

jcb

unread,
Jul 27, 2007, 3:53:07 PM7/27/07
to UsefulChem
Khalid - I understand that after a week some compounds are still not
here even though all were listed as next day ship? Can you double
check the info on the the website for the missing compounds?

On Jul 19, 12:21 pm, "Jean-Claude Bradley"


<jeanclaude.brad...@gmail.com> wrote:
> The order for the starting materials is going out today so we should find
> out soon :)
>

> On 7/19/07, Egon Willighagen <egon.willigha...@gmail.com> wrote:


>
>
>
>
>
>
>
> > On 7/18/07, Jean-Claude Bradley <jeanclaude.brad...@gmail.com> wrote:
> > > Here is a little summary of where we stand
> > >http://usefulchem.blogspot.com/2007/07/combiugi-time-for-synthesis.html
>
> > You're quite right. I should think before speak :) I did read that
> > item, but did not think about it when writing up my comment.
>
> > > We're going to make the top compounds that lend themselves to being made
> > > easily and cheaply. Hopefully there will be enough diversity there.
>
> > Still, I would guess that easily, cheap starting materials might still
> > give you enough flexibility...
>
> > Egon
>

> --
> Jean-Claude Bradley, Ph. D.
> E-Learning Coordinator for the College of Arts and Sciences

> Associate Professor of Chemistry
> Drexel University
>

> http://drexel-coas-elearning.blogspot.comhttp://drexel-coas-talks-mp3-podcast.blogspot.com/http://usefulchem.blogspot.com- Hide quoted text -
>
> - Show quoted text -

jcb

unread,
Jul 27, 2007, 4:21:39 PM7/27/07
to UsefulChem
Khalid tells me that we are only missing tosylmethyl isocyanide, which
is needed for most of the top 10 products. It turns out the ship date
has nothing to do with the arrival date. Maybe we can ask about
priority shipping next time.
Reply all
Reply to author
Forward
0 new messages