From owner-chemistry@ccl.net Mon Oct 19 11:32:00 2020 From: "HOUSTON Douglas DouglasR.Houston|*|ed.ac.uk" To: CCL Subject: CCL: Pharmacophore flexible search tool Message-Id: <-54189-201019101109-15205-ewLo7Rm734N7K6f3yEZRcA_-_server.ccl.net> X-Original-From: HOUSTON Douglas Content-Language: en-GB Content-Type: multipart/alternative; boundary="_000_DBAPR05MB7047554B786DCF44234387F5A81E0DBAPR05MB7047eurp_" Date: Mon, 19 Oct 2020 14:10:53 +0000 MIME-Version: 1.0 Sent to CCL by: HOUSTON Douglas [DouglasR.Houston{:}ed.ac.uk] --_000_DBAPR05MB7047554B786DCF44234387F5A81E0DBAPR05MB7047eurp_ Content-Type: text/plain; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable Hi Mark, Thanks for your detailed explanation and program suggestions. If adding ligand flexibility does not improve the results, then I guess I d= on't need it. But I also guess this is dependent on generating sufficient c= onformers. How many are sufficient? I realise this answer will depend on th= e number of rotatable bonds in each molecule ... Regards, Doug PS I can dock 100,000 molecules into a protein using e.g. PSOVina2 in a co= uple of days, and I expect small molecule comparison would go much faster b= ecause the number of grid points would be much lower. ________________________________ > From: owner-chemistry+douglasr.houston=3D=3Ded.ac.uk:+:ccl.net on behalf of Mark Mackey mark_._= cresset-group.com Sent: 15 October 2020 09:41 To: HOUSTON Douglas Subject: CCL: Pharmacophore flexible search tool Hi Doug, The main advantage of pharmacophore methods over more accurate 3D ligand si= milarity comparisons is speed: you can preprocess the database to generate = conformations and then searching that database is extremely fast. Once you = allow the method to change conformations you then have to be able to evalua= te the energy of those conformations (or else it will just find ridiculousl= y high-energy =93matches=94) and that is always going to be slow. Docking m= ethods are more-or-less forced to do this because they generally use very h= ard van der Waals interactions with the protein so that if a conformation i= s a quarter of an Angstrom out its docking score can increase significantly= . That=92s not true for pharmacophore methods, where in general minor confo= rmation change on a molecule matching the pharmacophore won=92t make it los= e the match. Even the more accurate 3D ligand similarity methods such as Phase shape, RO= CS or our shape/electrostatic similarity methods in Blaze and Forge (which,= quick plug, are free for graduate students to use!) tend to rely on precom= puted conformations as it is generally significantly more efficient and add= ing ligand flexibility does not improve the results. There are exceptions: FlexS from BioSolveIT and the eSim method from the Ja= in group both (I believe) incorporate a measure of ligand flexibility (the = latter only for fine-tuning after initially processing a conformer ensemble= ), but neither is free as far as I am aware. Is there a reason why you believe you need flexibility during the match its= elf? Regards, Mark -- Mark Mackey Chief Scientific Officer Cresset New Cambridge House, Bassingbourn Road, Litlington, Cambridgeshire, SG8 0SS= , UK tel: +44 (0)1223 858890 mobile: +44 (0)7595 099165 fax: +44 (0)1223 8= 53667 email: mark:+:cresset-group.com web: ww= w.cresset-group.com skype: mark_cresset > From: owner-chemistry+mark=3D=3Dcresset-bmd.com:+:ccl.net On Behalf Of HOUSTON Douglas DouglasR.= Houston[#]ed.ac.uk Sent: 15 October 2020 07:45 To: Mark Mackey Subject: CCL: Pharmacophore flexible search tool Hi Sergio, Thanks for your reply. Do you know the details of how ZINCPharmer compares = the molecules in the database with the pharmacophore? I note this from the = site: "The ZincPharmer search technology uses the Pharmer open source pharmacophore search technology to efficient= ly search a large database of fixed conformers for pharmacophore matches" There's a difference between creating a conformer library, and actually tre= ating each molecule as flexible. I already have ways of creating conformer libraries, but what results is a = coarse sampling of all possible conformations. What I'm looking for is a wa= y of treating the library compounds as actually flexible, in the same way t= hat e.g. Vina or Autodock treats the small molecule to be docked into a rec= eptor as flexible. Regards, Doug ________________________________ > From: owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-*-ccl.net on behalf of Sergio Vechi ve= chism/agmail.com Sent: 19 May 2020 21:59 To: HOUSTON Douglas Subject: CCL: Pharmacophore flexible search tool Dear Doug, You can build the pharmacophore model with PharmaGist (https://bioinfo3d.cs= .tau.ac.il/PharmaGist/php.php) and search Zinc database using ZINCPharmer (= http://zincpharmer.csb.pitt.edu/). Cheers, Sergio +--------------------------------------------------------------------------= --------------------------------------+ Dr. S=E9rgio Vechi | vechism:gmail.com Associate Professor Federal University of Alagoas | Campus Arapiraca | NCEx Av. Manuel Severino Barbosa, s/n, Bom Sucesso 57309-005 Arapiraca (AL) - Brazil +--------------------------------------------------------------------------= --------------------------------------+ On Tue, May 19, 2020 at 4:35 PM HOUSTON Douglas DouglasR.Houston]![ed.ac.uk= > wrote: Hi all, Does anyone know of a free (for academics at least) chemical search tool, t= hat will screen a large compound library (say an SDF of 100,000 molecules) = against a reference pharmacophore/ligand, and treat the library compounds a= s flexible and not rigid? I've asked this before but for some reason the formatting was so screwed up= I don't think most people could read it! Regards, Doug ---------------------------------------------------------------------------= --------------- Dr. Douglas R. Houston Senior Lecturer in Computational Biochemistry Institute of Quantitative Biology, Biochemistry and Biotechnology Room 2.12, Waddington 1 Building King's Buildings University of Edinburgh Edinburgh, EH9 3BF, UK Tel. 07986875743 The University of Edinburgh is a charitable body, registered in Scotland, w= ith registration number SC005336. --_000_DBAPR05MB7047554B786DCF44234387F5A81E0DBAPR05MB7047eurp_ Content-Type: text/html; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable
Hi Mark,

Thanks for your detailed explanation and program suggestions.

If adding ligand flexibility does not improve the results, then I gues= s I don't need it. But I also guess this is dependent on generating suffici= ent conformers. How many are sufficient? I realise this answer will depend = on the number of rotatable bonds in each molecule ...

Regards,
Doug

PS  I can dock 100,000 molecules into a protein using e.g. PSOVina2 in= a couple of days, and I expect small molecule comparison would go much fas= ter because the number of grid points would be much lower.


From: owner-chemistry+dougl= asr.houston=3D=3Ded.ac.uk:+:ccl.net <owner-chemistry+douglasr.houston=3D= =3Ded.ac.uk:+:ccl.net> on behalf of Mark Mackey mark_._cresset-group.com <owner-chemistry:+:ccl.net>
Sent: 15 October 2020 09:41
To: HOUSTON Douglas <DouglasR.Houston:+:ed.ac.uk>
Subject: CCL: Pharmacophore flexible search tool
 

Hi Doug,

 

The main advantage of pharmacophore methods over more accu= rate 3D ligand similarity comparisons is speed: you can preprocess the data= base to generate conformations and then searching that database is extremel= y fast. Once you allow the method to change conformations you then have to be able to evaluate the energy of= those conformations (or else it will just find ridiculously high-energy = =93matches=94) and that is always going to be slow. Docking methods are mor= e-or-less forced to do this because they generally use very hard van der Waals interactions with the protein so tha= t if a conformation is a quarter of an Angstrom out its docking score can i= ncrease significantly. That=92s not true for pharmacophore methods, where i= n general minor conformation change on a molecule matching the pharmacophore won=92t make it lose the match.

 

Even the more accurate 3D ligand similarity methods such a= s Phase shape, ROCS or our shape/electrostatic similarity methods in Blaze = and Forge (which, quick plug, are free for graduate students to use!) tend = to rely on precomputed conformations as it is generally significantly more efficient and adding ligand flexibil= ity does not improve the results.

 

There are exceptions: FlexS from BioSolveIT and the eSim m= ethod from the Jain group both (I believe) incorporate a measure of ligand = flexibility (the latter only for fine-tuning after initially processing a c= onformer ensemble), but neither is free as far as I am aware.

 

Is there a reason why you believe you need flexibility dur= ing the match itself?

 

Regards,

Mark

 

-- =

Mark Macke= y

Chief Scie= ntific Officer

Cresset

New Cambridge= House, Bassingbourn Road, Litlington, Cambridgeshire, SG8 0SS, UK

tel: +44 (0)1= 223 858890    mobile: +44 (0)7595 099165  &nb= sp; fax: +44 (0)1223 853667

email:= mark:+:= cresset-group.com    web: www.cresset-group.com&= nbsp;   skype: mark_cresset

 

 

 

From: owner-chemist= ry+mark=3D=3Dcresset-bmd.com:+:ccl.net <owner-chemistry+mark=3D=3Dcresse= t-bmd.com:+:ccl.net> On Behalf Of HOUSTON Douglas DouglasR.Houston[#]ed.ac.uk
Sent: 15 October 2020 07:45
To: Mark Mackey <mark:+:cresset-group.com>
Subject: CCL: Pharmacophore flexible search tool

 

Hi Sergio,

 

Thanks for your reply. Do you= know the details of how ZINCPharmer compares the molecules in the database= with the pharmacophore? I note this from the site:

 

"The ZincPharmer search technology uses the Pharmer open source pharmacophore search technology to efficiently search a large datab= ase of fixed conformers for pharmacophore matches"

 

There's a d= ifference between creating a conformer library, and actually treating each = molecule as flexible.<= /span>

 

I already h= ave ways of creating conformer libraries, but what results is a coarse samp= ling of all possible conformations. What I'm looking for is a way of treati= ng the library compounds as actually flexible, in the same way that e.g. Vina or Autodock treats= the small molecule to be docked into a receptor as flexible.

 

Regards,

Doug=

 

 

From:= owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-*-ccl.net <owner-chemist= ry+douglasr.houston=3D=3Ded.ac.uk-*-ccl.net> on behalf of Sergio Vechi v= echism/agmail.com <owner-chemistry-*-ccl.net>
Sent: 19 May 2020 21:59
To: HOUSTON Douglas <DouglasR.Houston-*-ed.ac.uk>
Subject: CCL: Pharmacophore flexible search tool

 

Dear Doug,

 

You can build the pharm= acophore model with PharmaGist (https://bioinfo3d.cs.tau.ac.il/PharmaGist/php.php) a= nd search Zinc database using ZINCPharmer (http://zincpharmer.csb.pitt.edu/).

 

Cheers,

 

Sergio

+----------------------------------------------= ------------------------------------------------------------------+
   Dr. S=E9rgio Vechi | vechism:gmail.com           =                     &nbs= p;  
     Associate Professor
     Federal University of Alagoas | Campus Arapiraca |= NCEx

     Av. Manuel Severino Barbosa= , s/n, Bom Sucesso

     57309-005 Arapiraca (A= L) - Brazil                  &= nbsp;     
+--------------------------------------------------------------------------= --------------------------------------+

 

 

On Tue, May 19, 2020 at 4:35 PM HOUSTON Douglas DouglasR.Houston]![ed.ac.uk <owner-chemistry|,|ccl.net> wrote:

Hi all,

 

Does anyone know of a free (for academics at least) chemica= l search tool, that will screen a large compound library (say an SDF o= f 100,000 molecules) against a reference pharmacophore/ligand, and treat the library compounds as flexible and not rigid?

 

I've asked this before but for some reason the formatting was so= screwed up I don't think most people could read it!

 

Regards,

Doug=

 

----------------------------------------------------------------= --------------------------

Dr. Douglas= R. Houston

Senior Lect= urer in Computational Biochemistry

Institute o= f Quantitative Biology, Biochemistry and Biotechnology

Room 2.12,&= nbsp;Waddington 1 Building

King's Buil= dings

University = of Edinburgh

Edinburgh,&= nbsp;EH9 3BF, UK

Tel. 079868= 75743

 

The University of Edinburgh is a charitable body, registered in Scotland, w= ith registration number SC005336.

--_000_DBAPR05MB7047554B786DCF44234387F5A81E0DBAPR05MB7047eurp_-- From owner-chemistry@ccl.net Mon Oct 19 16:28:00 2020 From: "Mark Mackey mark .. cresset-group.com" To: CCL Subject: CCL: Pharmacophore flexible search tool Message-Id: <-54190-201019145218-15868-uOoOD51suN1oDCLbTydlDw!A!server.ccl.net> X-Original-From: Mark Mackey Content-Language: en-US Content-Type: multipart/alternative; boundary="_000_DBBPR08MB4235ABFBA76DD6393D646D37971E0DBBPR08MB4235eurp_" Date: Mon, 19 Oct 2020 18:52:04 +0000 MIME-Version: 1.0 Sent to CCL by: Mark Mackey [mark*cresset-group.com] --_000_DBBPR08MB4235ABFBA76DD6393D646D37971E0DBBPR08MB4235eurp_ Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi Doug, There are a number of papers in the literature that suggest that you only n= eed surprisingly few conformations (eg https://link.springer.com/article/10= .1007%2Fs10822-014-9813-4). The argument is that as you increase the covera= ge of conformation space you make it more likely that you'll pick up a true= positive, but you also make it more likely that a false positive will give= a good score, so overall your hit rate doesn't improve. I think that there= 's definitely something to that claim, but I would argue that the threshold= of 10 confs or so that people tend to suggest is too low and the studies s= howing that this is sufficient are likely to be biased by all of the well-k= nown problems with retrospective virtual screening experiments (see https:/= /www.cresset-group.com/about/news/virtual-screening-how-many-conformations-= is-enough/). All of this is assuming that your compounds fall into the small-molecule-dr= uglike category, with a reasonable molecular weight and <10 rotatable bonds= . Within these constraints, 100 confs from a decent conformer generator is = probably describing the conformation space pretty well: our conformation hu= nter will get virtually all compounds with 6-7 rotatable bonds right (in th= e sense of having a conformation within 1-1.5A RMSD off the bioactive confo= rmation) with a 100 conf limit. That does fall off fairly quickly with flex= ibility, though: at 9-10 rotatable bonds you're looking at 50/50 whether yo= u'll find something close to the bioactive conf, and once you're much above= that the success rate drops off a cliff. Other conf hunt methods get simil= ar performance. If you are actually interested in a bunch of decapeptides or similar then i= t's basically impossible to pre-enumerate or describe the conformation spac= e. In that case some sort of flexible fit algorithm may be your only chance= of getting a sensible result. The caveat, of course, is that the average d= ecapeptide is going to match every single pharmacophore model you can think= of... The other problem with flexible fitting is accounting for the relative conf= ormation energy of the ligand (which is a much more complicated question th= an it sounds, but that's another post). If you pre-enumerate you can compar= e energies for each conformer to the energy of the "best" one that you foun= d and filter out the high-energy ones. With a flexible alignment you do nee= d to account for conformer energy (otherwise you'll just bend the molecule = into a dumb conformation to make it fit the pharmacophore), but you don't h= ave a known reference low-energy conformation to compare to so you won't kn= ow if your fitted ensemble are energetically reasonable or not. There are w= ays around that, but they aren't as rigorous. You are right that flexible fitting of ~100K mols to a pharmacophore or sha= pe model is perfectly doable in a few CPU days. Most pharmacophore software= has been designed with the aim of screening millions of molecules in a few= hours, however, which is why they tend to use pre-enumerated conformations= . Regards, Mark -- Mark Mackey Chief Scientific Officer Cresset New Cambridge House, Bassingbourn Road, Litlington, Cambridgeshire, SG8 0SS= , UK tel: +44 (0)1223 858890 mobile: +44 (0)7595 099165 fax: +44 (0)1223 8= 53667 email: mark:+:cresset-group.com web: www.cr= esset-group.com skype: mark_cresset > From: owner-chemistry+mark=3D=3Dcresset-bmd.com:+:ccl.net On Behalf Of HOUSTON Douglas DouglasR.Hous= ton|*|ed.ac.uk Sent: 19 October 2020 15:11 To: Mark Mackey Subject: CCL: Pharmacophore flexible search tool Hi Mark, Thanks for your detailed explanation and program suggestions. If adding ligand flexibility does not improve the results, then I guess I d= on't need it. But I also guess this is dependent on generating sufficient c= onformers. How many are sufficient? I realise this answer will depend on th= e number of rotatable bonds in each molecule ... Regards, Doug PS I can dock 100,000 molecules into a protein using e.g. PSOVina2 in a co= uple of days, and I expect small molecule comparison would go much faster b= ecause the number of grid points would be much lower. ________________________________ > From: owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-.-ccl.net on behalf of Mark Mackey mar= k_._cresset-group.com Sent: 15 October 2020 09:41 To: HOUSTON Douglas Subject: CCL: Pharmacophore flexible search tool Hi Doug, The main advantage of pharmacophore methods over more accurate 3D ligand si= milarity comparisons is speed: you can preprocess the database to generate = conformations and then searching that database is extremely fast. Once you = allow the method to change conformations you then have to be able to evalua= te the energy of those conformations (or else it will just find ridiculousl= y high-energy "matches") and that is always going to be slow. Docking metho= ds are more-or-less forced to do this because they generally use very hard = van der Waals interactions with the protein so that if a conformation is a = quarter of an Angstrom out its docking score can increase significantly. Th= at's not true for pharmacophore methods, where in general minor conformatio= n change on a molecule matching the pharmacophore won't make it lose the ma= tch. Even the more accurate 3D ligand similarity methods such as Phase shape, RO= CS or our shape/electrostatic similarity methods in Blaze and Forge (which,= quick plug, are free for graduate students to use!) tend to rely on precom= puted conformations as it is generally significantly more efficient and add= ing ligand flexibility does not improve the results. There are exceptions: FlexS from BioSolveIT and the eSim method from the Ja= in group both (I believe) incorporate a measure of ligand flexibility (the = latter only for fine-tuning after initially processing a conformer ensemble= ), but neither is free as far as I am aware. Is there a reason why you believe you need flexibility during the match its= elf? Regards, Mark -- Mark Mackey Chief Scientific Officer Cresset New Cambridge House, Bassingbourn Road, Litlington, Cambridgeshire, SG8 0SS= , UK tel: +44 (0)1223 858890 mobile: +44 (0)7595 099165 fax: +44 (0)1223 8= 53667 email: mark:+:cresset-group.com web: ww= w.cresset-group.com skype: mark_cresset > From: owner-chemistry+mark=3D=3Dcresset-bmd.com:+:ccl.net On Behalf Of HOUSTON Douglas DouglasR.= Houston[#]ed.ac.uk Sent: 15 October 2020 07:45 To: Mark Mackey Subject: CCL: Pharmacophore flexible search tool Hi Sergio, Thanks for your reply. Do you know the details of how ZINCPharmer compares = the molecules in the database with the pharmacophore? I note this from the = site: "The ZincPharmer search technology uses the Pharmer open source pharmacophore search technology to efficient= ly search a large database of fixed conformers for pharmacophore matches" There's a difference between creating a conformer library, and actually tre= ating each molecule as flexible. I already have ways of creating conformer libraries, but what results is a = coarse sampling of all possible conformations. What I'm looking for is a wa= y of treating the library compounds as actually flexible, in the same way t= hat e.g. Vina or Autodock treats the small molecule to be docked into a rec= eptor as flexible. Regards, Doug ________________________________ > From: owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-*-ccl.net on behalf of Sergio Vechi ve= chism/agmail.com Sent: 19 May 2020 21:59 To: HOUSTON Douglas Subject: CCL: Pharmacophore flexible search tool Dear Doug, You can build the pharmacophore model with PharmaGist (https://bioinfo3d.cs= .tau.ac.il/PharmaGist/php.php) and search Zinc database using ZINCPharmer (= http://zincpharmer.csb.pitt.edu/). Cheers, Sergio +--------------------------------------------------------------------------= --------------------------------------+ Dr. S=E9rgio Vechi | vechism:gmail.com Associate Professor Federal University of Alagoas | Campus Arapiraca | NCEx Av. Manuel Severino Barbosa, s/n, Bom Sucesso 57309-005 Arapiraca (AL) - Brazil +--------------------------------------------------------------------------= --------------------------------------+ On Tue, May 19, 2020 at 4:35 PM HOUSTON Douglas DouglasR.Houston]![ed.ac.uk= > wrote: Hi all, Does anyone know of a free (for academics at least) chemical search tool, t= hat will screen a large compound library (say an SDF of 100,000 molecules) = against a reference pharmacophore/ligand, and treat the library compounds a= s flexible and not rigid? I've asked this before but for some reason the formatting was so screwed up= I don't think most people could read it! Regards, Doug ---------------------------------------------------------------------------= --------------- Dr. Douglas R. Houston Senior Lecturer in Computational Biochemistry Institute of Quantitative Biology, Biochemistry and Biotechnology Room 2.12, Waddington 1 Building King's Buildings University of Edinburgh Edinburgh, EH9 3BF, UK Tel. 07986875743 The University of Edinburgh is a charitable body, registered in Scotland, w= ith registration number SC005336. --_000_DBBPR08MB4235ABFBA76DD6393D646D37971E0DBBPR08MB4235eurp_ Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable

Hi Doug,<= o:p>

&nbs= p;

There are= a number of papers in the literature that suggest that you only need surpr= isingly few conformations (eg h= ttps://link.springer.com/article/10.1007%2Fs10822-014-9813-4). The argu= ment is that as you increase the coverage of conformation space you make it= more likely that you’ll pick up a true positive, but you also make it more likely that a false positive will= give a good score, so overall your hit rate doesn’t improve. I think= that there’s definitely something to that claim, but I would argue t= hat the threshold of 10 confs or so that people tend to suggest is too low and the studies showing that this is sufficient= are likely to be biased by all of the well-known problems with retrospecti= ve virtual screening experiments (see https://www.cresset-group.com/about/news/virtual-screening-how-many-conform= ations-is-enough/).

&nbs= p;

All of th= is is assuming that your compounds fall into the small-molecule-druglike ca= tegory, with a reasonable molecular weight and <10 rotatable bonds. With= in these constraints, 100 confs from a decent conformer generator is probably describing the conformation space p= retty well: our conformation hunter will get virtually all compounds with 6= -7 rotatable bonds right (in the sense of having a conformation within 1-1.= 5A RMSD off the bioactive conformation) with a 100 conf limit. That does fall off fairly quickly with flexibility,= though: at 9-10 rotatable bonds you’re looking at 50/50 whether you&= #8217;ll find something close to the bioactive conf, and once you’re = much above that the success rate drops off a cliff. Other conf hunt methods get similar performance.

&nbs= p;

If you ar= e actually interested in a bunch of decapeptides or similar then it’s= basically impossible to pre-enumerate or describe the conformation space. = In that case some sort of flexible fit algorithm may be your only chance of getting a sensible result. The caveat, of cours= e, is that the average decapeptide is going to match every single pharmacop= hore model you can think of…

&nbs= p;

The other= problem with flexible fitting is accounting for the relative conformation = energy of the ligand (which is a much more complicated question than it sou= nds, but that’s another post). If you pre-enumerate you can compare energies for each conformer to the energy of= the “best” one that you found and filter out the high-energy o= nes. With a flexible alignment you do need to account for conformer energy = (otherwise you’ll just bend the molecule into a dumb conformation to make it fit the pharmacophore), but you don’t= have a known reference low-energy conformation to compare to so you won= 217;t know if your fitted ensemble are energetically reasonable or not. The= re are ways around that, but they aren’t as rigorous.

&nbs= p;

You are r= ight that flexible fitting of ~100K mols to a pharmacophore or shape model = is perfectly doable in a few CPU days. Most pharmacophore software has been= designed with the aim of screening millions of molecules in a few hours, however, which is why they tend to u= se pre-enumerated conformations.

 

Regards,

Mark

 

--

Mark Mackey

Chief Scientific Officer

Cresset

New Cambridge House, Bassingbourn Road, Litlington, Cambridgeshire= , SG8 0SS, UK

tel: +44 (0)1223 858890    mobile: +44 (0)7595= 099165    fax: +44 (0)1223 853667

email: mark:+:cress= et-group.com    web: www.cresset-group.com&n= bsp;   skype: mark_cresset

 

&nbs= p;

&nbs= p;

From: owner-chemistry+mark=3D=3Dcresset-bmd.com:+:ccl.net <owner-che= mistry+mark=3D=3Dcresset-bmd.com:+:ccl.net> On Behalf Of HOUSTON Douglas DouglasR.Houston|*|ed.ac.uk
Sent: 19 October 2020 15:11
To: Mark Mackey <mark:+:cresset-group.com>
Subject: CCL: Pharmacophore flexible search tool

 

Hi Mark= ,

&n= bsp;

Thanks = for your detailed explanation and program suggestions.

&n= bsp;

If = ;adding ligand flexibility does not improve the results, then I guess I don= 't need it. But I also guess this is dependent on generating sufficient con= formers. How many are sufficient? I realise this answer will depend on the number of rotatable bonds in each molecule = ...

&n= bsp;

Regards= ,

Doug

&n= bsp;

PS = ; I can dock 100,000 molecules into a protein using e.g. PSOVina2 in a coup= le of days, and I expect small molecule comparison would go much faster bec= ause the number of grid points would be much lower.

 

From: owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-.-cc= l.net <owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-.-ccl.net> on b= ehalf of Mark Mackey mark_._cresset-group.com <owner-chemistry-.-ccl.net= >
Sent: 15 October 2020 09:41
To: HOUSTON Douglas <DouglasR.Houston-.-ed.ac.uk>
Subject: CCL: Pharmacophore flexible search tool <= /p>

 

Hi Doug,

 

The main advantage of pharmacophore methods over mo= re accurate 3D ligand similarity comparisons is speed: you can preprocess t= he database to generate conformations and then searching that database is e= xtremely fast. Once you allow the method to change conformations you then have to be able to evaluate the en= ergy of those conformations (or else it will just find ridiculously high-en= ergy “matches”) and that is always going to be slow. Docking me= thods are more-or-less forced to do this because they generally use very hard van der Waals interactions with the protein s= o that if a conformation is a quarter of an Angstrom out its docking score = can increase significantly. That’s not true for pharmacophore methods= , where in general minor conformation change on a molecule matching the pharmacophore won’t make it lose t= he match.

 

Even the more accurate 3D ligand similarity methods= such as Phase shape, ROCS or our shape/electrostatic similarity methods in= Blaze and Forge (which, quick plug, are free for graduate students to use!= ) tend to rely on precomputed conformations as it is generally significantly more efficient and adding ligand flexibil= ity does not improve the results.

 

There are exceptions: FlexS from BioSolveIT and the= eSim method from the Jain group both (I believe) incorporate a measure of = ligand flexibility (the latter only for fine-tuning after initially process= ing a conformer ensemble), but neither is free as far as I am aware.

 

Is there a reason why you believe you need flexibil= ity during the match itself?

 

Regards,

Mark

 

--

Mark Mackey

Chief Scientific Officer

Cresset

New Cambridge House, Bassingbourn Road, Litlington, Cambridgeshir= e, SG8 0SS, UK

tel: +44 (0)1223 858890    mobile: +44 (0)759= 5 099165    fax: +44 (0)1223 853667

email: mark:+:c= resset-group.com    web: www.cresset-group.com&n= bsp;   skype: mark_cresset

 

 

 

From: owner-chemistry+mark=3D=3Dcresset-bmd.com:+:ccl.net <owner-c= hemistry+mark=3D=3Dcresset-bmd.com:+:ccl.net> On Behalf Of HOUSTON Douglas DouglasR.Houston[#]ed.ac.uk
Sent: 15 October 2020 07:45
To: Mark Mackey <mark:+:cresset-group.com>
Subject: CCL: Pharmacophore flexible search tool

 

Hi Ser= gio,

 =

Thanks= for your reply. Do you know the details of how ZINCPharmer compares the mo= lecules in the database with the pharmacophore? I note this from the site:<= /span>

 =

"= The ZincPharmer search technology uses t= he Pharmer open source pharmacophore search technology to efficiently search a large datab= ase of fixed conformers for pharmacophore matches"

 =

There's a difference between creating a conformer library, and a= ctually treating each molecule as flexible.

 =

I already have ways of creating conformer libraries, but what re= sults is a coarse sampling of all possible conformations. What I'm looking = for is a way of treating the library compounds as actually flexible, in the same way that e.g. Vina or A= utodock treats the small molecule to be docked into a receptor as flexible.=

 =

Regards,

Doug

 

 =

From: owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-*-c= cl.net <owner-chemistry+douglasr.houston=3D=3Ded.ac.uk-*-ccl.net> on = behalf of Sergio Vechi vechism/agmail.com <owner-chemistry-*-ccl.net>=
Sent: 19 May 2020 21:59
To: HOUSTON Douglas <DouglasR.Houston-*-ed.ac.uk>
Subject: CCL: Pharmacophore flexible search tool <= /p>

 

Dear Doug,

 

You can build the pharmacophore model with PharmaGist (https://bioinfo3d.cs.tau.ac.il/P= harmaGist/php.php) and search Zinc database using ZINCPharmer (http://zin= cpharmer.csb.pitt.edu/).

 

Cheers,

 

Sergio

+----------------------= ---------------------------------------------------------------------------= ---------------+
   Dr. S=E9rgio Vechi | vechism:gmail.com           =                     &nbs= p;  
     Associate Professor
     Federal University of Alagoas | Campus Arapiraca |= NCEx

     Av.= Manuel Severino Barbosa, s/n, Bom Sucesso

    = ; 57309-005 Arapiraca (AL) - Brazil           = ;             
+--------------------------------------------------------------------------= --------------------------------------+

 

 

On Tue, May 19, 2020 at 4:35 PM HOUSTON Douglas Dou= glasR.Houston]![ed.ac.uk <owner-chemistry|,|ccl.net> wrote:

Hi all,=

 =

Does anyone know of a fre= e (for academics at least) chemical search tool, that will screen= a large compound library (say an SDF of 100,000 molecules) against a reference pharmacophore/ligand, and treat the library compounds = as flexible and not rigid?

 =

I've asked this before bu= t for some reason the formatting was so screwed up I don't think most peopl= e could read it!

 =

Regards,

Doug

 =

-------------------------= -----------------------------------------------------------------

Dr. Douglas R. Houston

Senior Lecturer in Computational Biochemistry<= /p>

Institute of Quantitative Biology, Biochemistry and Biotechnolog= y

Room 2.12, Waddington 1 Building

King's Buildings

University of Edinburgh

Edinburgh, EH9 3BF, UK

Tel. 07986875743

 =

The University of Edinburgh is a charitable body, r= egistered in Scotland, with registration number SC005336.

--_000_DBBPR08MB4235ABFBA76DD6393D646D37971E0DBBPR08MB4235eurp_--