From owner-chemistry@ccl.net Thu Jul 9 02:47:01 2009 From: "Andreas Klamt klamt(a)cosmologic.de" To: CCL Subject: CCL: Question on sterol/lipid interactions and CCL Message-Id: <-39718-090709023346-15032-kG5VhhWi3X2YezC8xLlChw*|*server.ccl.net> X-Original-From: Andreas Klamt Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-1; format=flowed Date: Thu, 09 Jul 2009 08:33:23 +0200 MIME-Version: 1.0 Sent to CCL by: Andreas Klamt [klamt:-:cosmologic.de] Hi David, these are a lot and very difficult questions and I am far from being=20 able to answer them. I just like to bring our COSMOmic method to your=20 attention. It allows you to do simulate the positional, orientational=20 and conformational preference of molecules in the presence of an=20 effective background description of the biomembrane (or any kind of=20 micelle). The description of the solute can be done with quantum=20 chemistry (DFT) and is thus independent of force fields. This may help=20 answer your questions. What COSMOmic currently cannot do automatically=20 is to construct and predict the details of the micelle composition and=20 geometry based on the input of the chemical compounds, (but we have=20 ideas in this directio which we might share and exploit together with=20 some interested and membrane modelling experienced partner). Perhaps have a look to pubs.acs.org/doi/abs/10.1021/jp801736k and/or http://www.cosmologic.de/index.php?cosId=3D2130&crId=3D2 Andreas David Mannock dmannock{=3D}ualberta.ca schrieb: > Hi! I do experimental work on lipid/sterol mixtures and have been=20 > looking at the existing computer calculations and bilayer simulations=20 > of these mixtures in an effort to interpret thermodynamic and=20 > structural data. It seems to me that there are a lot of problems in=20 > this area. Different molecular force fields give disparate pictures,=20 > electrostatic calculations are performed independently and are not=20 > part of the conformational calculations. Where steepest descent=20 > algorithms are employed in conformational analysis, they obtain the=20 > global minimum energy conformation and those conformations are used in = > simulations, yet they may be only one of several conformations in a=20 > dynamic system even where there are lateral forces between molecules=20 > in the bilayer. There is some recent experimental and computer work=20 > which seems to acknowledge that the alkyl side-chain may be kinked and = > not fully extended in some biological sterols, but these are in the=20 > minority and explanations for such conformations are weak. Also,=20 > contrary to established belief, the all-/trans/ polycyclic ring system = > is not completely inflexible (this is even evident in crystal=20 > structures) and the conformations of rings A, B and D are not the same = > in all sterols. In fact, the addition of methyl groups, changes in=20 > functional group chemistry and the insertion of double bonds into the=20 > ring and side-chain introduce major changes in molecular properties.=20 > Why are these calculations not detecting this when the indications are = > there in other areas of the sterol and steroid literature? Why are the = > molecular electronic properties not included in many conformational=20 > analyses prior to the bilayer simulations and why are the global=20 > minimum energy conformations used in the simulations when it might be=20 > better to consider the dynamic electrostatic potential surfaces? There = > is some interesting calculations which look at the contact=20 > surfaces/fingerprints between phospholipids and sterols using MD=20 > methods. Is anyone doing this kind of analysis on sterols other than=20 > cholesterol? David Mannock --=20 PD. Dr. Andreas Klamt CEO / Gesch=E4ftsf=FChrer COSMOlogic GmbH & Co. KG Burscheider Strasse 515 D-51381 Leverkusen, Germany phone +49-2171-731681 fax +49-2171-731689 e-mail klamt(-)cosmologic.de web www.cosmologic.de HRA 20653 Landgericht Koeln, GF: Dr. Andreas Klamt Komplementaer: COSMOlogic Verwaltungs GmbH HRB 49501 Landgericht Koeln, GF: Dr. Andreas Klamt From owner-chemistry@ccl.net Thu Jul 9 04:39:01 2009 From: "Simon Cross simon]~[moldiscovery.com" To: CCL Subject: CCL:G: CBS-Q method in druglike molecules Message-Id: <-39719-090708151726-953-C++lnH8l5qwyPxxOCF/dOQ- -server.ccl.net> X-Original-From: Simon Cross Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=ISO-8859-1; format=flowed Date: Wed, 08 Jul 2009 19:21:16 +0100 MIME-Version: 1.0 Sent to CCL by: Simon Cross [simon!=!moldiscovery.com] Hi Pavle, have you considered other methods for pKa calculation of drug-like molecules? We recently published a method (Milletti, F. et al. (2007) New and Original pKa Prediction Method Using Grid Molecular Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is used in our software MoKa. You can see more about it at www.moldiscovery.com/soft_moka.php We get an RMSE of 0.4 log units for the training set of 26000 data points (very close to experimental error) and external prediction error of 0.7 log units. Simon Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote: > Sent to CCL by: "Pavle Mocilac" [pavle.mocilac2{:}mail.dcu.ie] > My dear colleagues, > > Recently I sent the question about the calculation of pKa of small druglike molecules. The method asks employing high accuracy models like G1, G2, CBS-Q in order to get accurate Free Gibbs energies using Gaussian 09, of course.. > Now, it seems to me that this model works OK, and fast enough for very small molecules up to 10 or 12 atoms (hydrogen included). But, my molecule has 25 atoms and currently is stacked at QCISD(T)/6-31+G(d') step, calculating something called "triples" for more than 2 days!!!!! > I would like to emphasize that this job is being calculating with Gaussian 09 (Linux version), using lots of memory () and 8 processors at SGI Altix ICE 8200EX. > What is the limit of CBS-Q method regarding the number of atoms? Is 25 atoms too much? Is there any option to use some kind of additional keywords to exclude "triples" calculation at QCISD(T)/6-31+G(d') step? Will that exclusion hamper accurate result? Is it better to use CBS-4M?> > > -- Simon Cross Snr Scientist & Product Manager Molecular Discovery From owner-chemistry@ccl.net Thu Jul 9 08:59:01 2009 From: "Pavle Mocilac pavle.mocilac2^-^mail.dcu.ie" To: CCL Subject: CCL:G: CBS-Q method in druglike molecules Message-Id: <-39720-090709081303-24704-rJgIw2zR/3KJ03w7m2ylcQ(_)server.ccl.net> X-Original-From: Pavle Mocilac Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=iso-8859-1 Date: Thu, 9 Jul 2009 13:12:28 +0100 MIME-Version: 1.0 Sent to CCL by: Pavle Mocilac [pavle.mocilac2===mail.dcu.ie] Dear Simon, Thank you for the information about this software, but no, I have not and will not consider any other methods beside those that can be made with Gaussian 03 or 09 and described in "CaballeroNA, Melendez FJ, Munoz-Cara C, et al. Biophys. Chem 124 (2) 155-160". We do not have any intention to buy any new additional software but to use the one we have. Is seems to me that there are numerous people on this CCL that behave as salesman and always want to solve my problem by selling me some additional software. The reason we are using Gaussian is deeper, and for the sake of consistency and comparability we do not have attention to use so many different programmes that works in different way. I would like to repeat that CBS-Q is my main problem and therefore I will be very happy and very grateful if there is somebody in this CCL who can help me with CBS-Q calculations in Gaussian. Best regards, Pavle Mocilac ============================================ Pavle Mocilac Postgraduate Researcher T3 - Targeted Therapeutics and Theranostics Room X249, School of Chemistry Dublin City University Dublin 9, Dublin, Ireland mobile: +353872167022 mailto:pavle.mocilac2+/-mail.dcu.ie ============================================ Wednesday, July 8, 2009, 7:21:16 PM, you wrote: > Sent to CCL by: Simon Cross [simon!=!moldiscovery.com] > Hi Pavle, have you considered other methods for pKa calculation of > drug-like molecules? We recently published a method (Milletti, F. et al. > (2007) New and Original pKa Prediction Method Using Grid Molecular > Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is used > in our software MoKa. > You can see more about it at www.moldiscovery.com/soft_moka.php > We get an RMSE of 0.4 log units for the training set of 26000 data > points (very close to experimental error) and external prediction error > of 0.7 log units. > Simon > Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote: >> Sent to CCL by: "Pavle Mocilac" [pavle.mocilac2{:}mail.dcu.ie] >> My dear colleagues, >> >> Recently I sent the question about the calculation of pKa of small druglike molecules. The method asks employing high accuracy models like G1, G2, CBS-Q in order to get accurate Free Gibbs energies using Gaussian 09, of course.. >> Now, it seems to me that this model works OK, and fast enough for very small molecules up to 10 or 12 atoms (hydrogen included). But, my molecule has 25 atoms and currently is stacked at QCISD(T)/6-31+G(d') step, calculating something called "triples" for more than 2 days!!!!! >> I would like to emphasize that this job is being calculating with Gaussian 09 (Linux version), using lots of memory () and 8 processors at SGI Altix ICE 8200EX. >> What is the limit of CBS-Q method regarding the number of atoms? Is 25 atoms too much? Is there any option to use some kind of additional keywords to exclude "triples" calculation at QCISD(T)/6-31+G(d') step? Will that exclusion hamper accurate result? Is it better to use CBS-4M?> >> >> From owner-chemistry@ccl.net Thu Jul 9 09:53:01 2009 From: "Simon Cross simon],[moldiscovery.com" To: CCL Subject: CCL:G: CBS-Q method in druglike molecules Message-Id: <-39721-090709095139-26833-ZThP2BQfYlJGtkpeJ/XzLA_+_server.ccl.net> X-Original-From: Simon Cross Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=ISO-8859-1; format=flowed Date: Thu, 09 Jul 2009 14:51:43 +0100 MIME-Version: 1.0 Sent to CCL by: Simon Cross [simon,moldiscovery.com] Dear Pavle, of course you can choose the method you like, I only responded because it seemed you were getting frustrated by spending days on the calculation and our method is interactive and I know for a fact that it is interesting to many people. I would be grateful if you didn't take your frustration out on me when I was only trying to help you. Simon Pavle Mocilac pavle.mocilac2^-^mail.dcu.ie wrote: > Sent to CCL by: Pavle Mocilac [pavle.mocilac2===mail.dcu.ie] > Dear Simon, > > Thank you for the information about this software, but no, I have not and will not consider any other methods beside those that can be made with Gaussian 03 or 09 and described in "CaballeroNA, Melendez FJ, Munoz-Cara C, et al. Biophys. Chem 124 (2) > 155-160". > We do not have any intention to buy any new additional software but to use the one we have. Is seems to me that there are numerous people on this CCL that behave as salesman and always want to solve my problem by selling > me some additional software. > The reason we are using Gaussian is deeper, and for the sake of consistency and comparability we do not have attention to use so many different programmes that works in different way. > I would like to repeat that CBS-Q is my main problem and therefore I will be very happy and very grateful if there is somebody in this CCL who can help me with CBS-Q calculations in Gaussian. > > > Best regards, > > Pavle Mocilac > > ============================================ > Pavle Mocilac > Postgraduate Researcher > T3 - Targeted Therapeutics and Theranostics > Room X249, School of Chemistry > Dublin City University > Dublin 9, Dublin, Ireland > mobile: +353872167022 > mailto:pavle.mocilac2~!~mail.dcu.ie > ============================================ > Wednesday, July 8, 2009, 7:21:16 PM, you wrote: > > > >> Sent to CCL by: Simon Cross [simon!=!moldiscovery.com] >> > > >> Hi Pavle, have you considered other methods for pKa calculation of >> drug-like molecules? We recently published a method (Milletti, F. et al. >> (2007) New and Original pKa Prediction Method Using Grid Molecular >> Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is used >> in our software MoKa. >> > > >> You can see more about it at www.moldiscovery.com/soft_moka.php >> > > >> We get an RMSE of 0.4 log units for the training set of 26000 data >> points (very close to experimental error) and external prediction error >> of 0.7 log units. >> > > >> Simon >> > > >> Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote: >> >>> Sent to CCL by: "Pavle Mocilac" [pavle.mocilac2{:}mail.dcu.ie] >>> My dear colleagues, >>> >>> Recently I sent the question about the calculation of pKa of small druglike molecules. The method asks employing high accuracy models like G1, G2, CBS-Q in order to get accurate Free Gibbs energies using Gaussian 09, of course.. >>> Now, it seems to me that this model works OK, and fast enough for very small molecules up to 10 or 12 atoms (hydrogen included). But, my molecule has 25 atoms and currently is stacked at QCISD(T)/6-31+G(d') step, calculating something called "triples" for more than 2 days!!!!! >>> I would like to emphasize that this job is being calculating with Gaussian 09 (Linux version), using lots of memory () and 8 processors at SGI Altix ICE 8200EX. >>> What is the limit of CBS-Q method regarding the number of atoms? Is 25 atoms too much? Is there any option to use some kind of additional keywords to exclude "triples" calculation at QCISD(T)/6-31+G(d') step? Will that exclusion hamper accurate result? Is it better to use CBS-4M?> > > -- Simon Cross Snr Scientist & Product Manager Molecular Discovery From owner-chemistry@ccl.net Thu Jul 9 10:27:01 2009 From: "VITORGE Pierre 094605 Pierre.VITORGE%a%cea.fr" To: CCL Subject: CCL: pKa calculation of strong acid Message-Id: <-39722-090709073219-19699-ARmof1Uer7tYW3WEbq6v/g|-|server.ccl.net> X-Original-From: "VITORGE Pierre 094605" Content-class: urn:content-classes:message Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="iso-8859-1" Date: Thu, 9 Jul 2009 11:44:50 +0200 MIME-Version: 1.0 Sent to CCL by: "VITORGE Pierre 094605" [Pierre.VITORGE*cea.fr] Dave, pKa =3D -6 in aqueous solution, means it cannot be measured. We are both = right: the value cannot be measured, but it might very well be the good = value, which I am explaining below. pKa =3D -lgKa is a definition of pKa, where=20 Ka =3D [H+] [Cl-] / [HCl] [X] is the concentration of species X, here X is an aqueous specie (H+, = Cl- or HCl). For simplicity I am using concentrations meaning the = (small) activity corrections are included in Ka. A first (chemical) meaning of pKa is (pH) half point reaction, namely = for 50% dissociation [Cl-]_1/2 =3D [HCl]_1/2, where subscript _1/2 is to = stress concentrations are for special (chemical) conditions. = Consequently Ka =3D [H+]_1/2 equivalently pKa =3D -lg[H+]_1/2 and pKa=B0 =3D pH_1/2 a well known formula in solution chemistry, where = superscript=B0 is for standard conditions (infinite dilution where = -lg[H+] -> pH) In this respect pKa of strong acid has no finite value, since strong = acid means completely dissociated in any chemical conditions, namely = [HCl]=3D0; which is indeed not possible for finite values of Ka. However, Ka can directly be measured only when the dissociated form can = be detected, typically if the detection limit is 1% only pKa > -2 can be = directly measured, and "no finite value" actually means "> -2". (Partial) conclusion: do not write pKa of strong acid; it is chocking = for those who currently measure or use pKas knowing the meanings of the = concept. Another meaning of pKa is form thermodynamics: delta_rG =3D -RTlnK here = applied to K =3D Ka. You can very well calculate delta_rGa =3D G(H+) + G(Cl-) - G(HCl) (again note they are aqueous species, especially not HCl(g)), where G(X) = can be taken from thermochemical data bases (as G(X) =3D delta_fG(X)) or = calculated by molecular modelling.=20 Now imagine delta_rGa have been estimated in this way giving pKa=3D-6. = This means that in the most favourable conditions to form HCl, typically = 10 mol.L-1 total HCl concentration, the concentrations of the actual = species in solution would be [H+] # 10 mol.L-1 [Cl-] # 10 mol.L-1 [HCl] =3D 10^-6 mol.L-1, which can probably not be detected, but might = very well exist. The reasoning still stand even for concentrations less = than Avogadro Number per litre, and the result is meaningful: the (Ka) = reaction can be used in thermodynamic cycles, typically with another = acid-base reaction to decide in which way the proton will be = transferred. Conclusion: what is your actual problem? Do you really need the pKa of = HCl in liquid water? Or is it only part of your calculations? In this = later case can it be cancelled out by using other calculation = strategies?=20 This was actually the origin of my remark. Pierre Vitorge -----Message d'origine----- De=A0: owner-chemistry+pierre.vitorge=3D=3Dcea.fr!=!ccl.net = [mailto:owner-chemistry+pierre.vitorge=3D=3Dcea.fr!=!ccl.net] De la part = de case case~!~biomaps.rutgers.edu Envoy=E9=A0: mercredi 8 juillet 2009 16:44 =C0=A0: VITORGE Pierre 094605 Objet=A0: CCL: pKa calculation of strong acid Sent to CCL by: case [case##biomaps.rutgers.edu] On Tue, Jul 07, 2009, VITORGE Pierre 094605 Pierre.VITORGE^^cea.fr = wrote: >=20 > The pKa of a strong acid have no finite value, namely it does not = exist. These values certainly exist. For the HCl example that was in the original post, the pKa in aqueous solution is around -6, although this seems to have a big uncertainty. You would have to spend some time to find the original determination -- a secondary reference to start from would be Albert and Serjeant, "The Determination of Ionization = Constants", Chapman & Hall, 1984. ....dave case -=3D This is automatically added to each message by the mailing script = =3D-http://www.ccl.net/cgi-bin/ccl/send_ccl_messageSubscribe/Unsubscribe:=20Job: http://www.ccl.net/jobs=20http://www.ccl.net/spammers.txt From owner-chemistry@ccl.net Thu Jul 9 11:02:00 2009 From: "James Robinson james.robinson+/-prosonix.co.uk" To: CCL Subject: CCL:G: CBS-Q pka calculation Message-Id: <-39723-090709104907-18341-lFz5rIvY5Yc2UhtpVY6OjQ[A]server.ccl.net> X-Original-From: "James Robinson" Date: Thu, 9 Jul 2009 10:49:03 -0400 Sent to CCL by: "James Robinson" [james.robinson_-_prosonix.co.uk] Dear Pavle, Your problem is related to hardware, i suspect. Remember that there may be options within gaussian to make use of more efficient memory, are you optmising the starting structures with a lower level of theory, are you using options like symmetry, are your calculations running as efficiently as possible? Would optimising strutures with lower level of theory help in getting the strutures for the calculation in better time. Are using ramen options in freq calculations. If you cannot improve the efficiency of your calculation, how can you improve your hardware, are you using 32 bit software on a 64 bit capable machine.. Regards James Prosonix Oxoford. From owner-chemistry@ccl.net Thu Jul 9 11:37:00 2009 From: "=?ISO-8859-1?Q?Gon=E7alo_C=2E_Justino?= jgcj^^fct.unl.pt" To: CCL Subject: CCL:G: CBS-Q method in druglike molecules Message-Id: <-39724-090709104734-16427-N7CdpzYbA2BowA1QP3ZusQ_-_server.ccl.net> X-Original-From: =?ISO-8859-1?Q?Gon=E7alo_C=2E_Justino?= Content-Type: multipart/alternative; boundary=0016e649d942099b62046e4682e3 Date: Thu, 9 Jul 2009 15:16:55 +0100 MIME-Version: 1.0 Sent to CCL by: =?ISO-8859-1?Q?Gon=E7alo_C=2E_Justino?= [jgcj**fct.unl.pt] --0016e649d942099b62046e4682e3 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable dear pavle, as far as i'm aware (someone correct me) the triples corection you refer to come from the '(T)' part in the QCISD(T) keyword. Using QCISD only (no T into it) would not require triples correction. can you afford to use the QCISD instead of the QCISD(T) ? I know this is the simple answer, but it's the only one I can come up with. the triples come from *J. Chem. Phys.*, *87* (1987) 5968-75.. hope it helps at all, gon=E7alo 2009/7/9 Pavle Mocilac pavle.mocilac2^-^mail.dcu.ie > > Sent to CCL by: Pavle Mocilac [pavle.mocilac2=3D=3D=3Dmail.dcu.ie] > Dear Simon, > > Thank you for the information about this software, but no, I have not and > will not consider any other methods beside those that can be made with > Gaussian 03 or 09 and described in "CaballeroNA, Melendez FJ, Munoz-Cara = C, > et al. Biophys. Chem 124 (2) > 155-160". > We do not have any intention to buy any new additional software but to us= e > the one we have. Is seems to me that there are numerous people on this CC= L > that behave as salesman and always want to solve my problem by selling > me some additional software. > The reason we are using Gaussian is deeper, and for the sake of consisten= cy > and comparability we do not have attention to use so many different > programmes that works in different way. > I would like to repeat that CBS-Q is my main problem and therefore I will > be very happy and very grateful if there is somebody in this CCL who can > help me with CBS-Q calculations in Gaussian. > > > Best regards, > > Pavle Mocilac > > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > Pavle Mocilac > Postgraduate Researcher > T3 - Targeted Therapeutics and Theranostics > Room X249, School of Chemistry > Dublin City University > Dublin 9, Dublin, Ireland > mobile: +353872167022 > mailto:pavle.mocilac2~!~mail.dcu.ie > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > Wednesday, July 8, 2009, 7:21:16 PM, you wrote: > > > > Sent to CCL by: Simon Cross [simon!=3D!moldiscovery.com] > > > Hi Pavle, have you considered other methods for pKa calculation of > > drug-like molecules? We recently published a method (Milletti, F. et al= . > > (2007) New and Original pKa Prediction Method Using Grid Molecular > > Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is used > > in our software MoKa. > > > You can see more about it at www.moldiscovery.com/soft_moka.php > > > We get an RMSE of 0.4 log units for the training set of 26000 data > > points (very close to experimental error) and external prediction error > > of 0.7 log units. > > > Simon > > > Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote: > >> Sent to CCL by: "Pavle Mocilac" [pavle.mocilac2{:}mail.dcu.ie] > >> My dear colleagues, > >> > >> Recently I sent the question about the calculation of pKa of small > druglike molecules. The method asks employing high accuracy models like G= 1, > G2, CBS-Q in order to get accurate Free Gibbs energies using Gaussian 09,= of > course.. > >> Now, it seems to me that this model works OK, and fast enough for very > small molecules up to 10 or 12 atoms (hydrogen included). But, my molecul= e > has 25 atoms and currently is stacked at QCISD(T)/6-31+G(d') step, > calculating something called "triples" for more than 2 days!!!!! > >> I would like to emphasize that this job is being calculating with > Gaussian 09 (Linux version), using lots of memory () and 8 processors at = SGI > Altix ICE 8200EX. > >> What is the limit of CBS-Q method regarding the number of atoms? Is 25 > atoms too much? Is there any option to use some kind of additional keywor= ds > to exclude "triples" calculation at QCISD(T)/6-31+G(d') step? Will that > exclusion hamper accurate result? Is it better to use CBS-4M?> > >> > >> > > > > -=3D This is automatically added to each message by the mailing script = =3D-> > > --0016e649d942099b62046e4682e3 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable dear pavle,

as far as i'm aware (someone correct me) the triples= corection you refer to come from the '(T)' part in the QCISD(T) ke= yword. Using QCISD only (no T into it) would not require triples correction= . can you afford to use the QCISD instead of the QCISD(T) ? I know this is = the simple answer, but it's the only one I can come up with.

the triples come from J. Chem. Phys., 87 (1987) 5968-75..=

hope it helps at all,
gon=E7alo




2009/7/9 Pavle Mocilac pavle.mocilac2^-^= mail.dcu.ie <owner-chemistry..ccl.net>=

Sent to CCL by: Pavle Mocilac [pavle.mocilac2=3D=3D=3Dmail.dcu.ie]
Dear Simon,

Thank you for the information about this software, but no, I have not and w= ill not consider any other methods beside those that can be made with Gauss= ian 03 or 09 and described in "CaballeroNA, Melendez FJ, Munoz-Cara C,= et al. =A0 Biophys. Chem =A0124 =A0(2)
155-160".
We do not have any intention to buy any new additional software but to use = the one we have. Is seems to me that there are numerous people on this CCL = =A0that behave as salesman and always want to solve my problem by selling me some additional software.
The reason we are using Gaussian is deeper, and for the sake of consistency= and comparability we do not have attention to use so many different progra= mmes that works in different way.
I would like to repeat that CBS-Q is my main problem and therefore I will b= e very happy and very grateful if there is somebody in this CCL who can hel= p me with CBS-Q calculations in Gaussian.


Best regards,

Pavle Mocilac

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
Pavle Mocilac
Postgraduate Researcher
T3 - Targeted Therapeutics and Theranostics
Room X249, School of Chemistry
Dublin City University
Dublin 9, Dublin, Ireland
mobile: +353872167022
mailto:pavle.mocilac2~!~mail.dcu.ie
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
Wednesday, July 8, 2009, 7:21:16 PM, you wrote:


> Sent to CCL by: Simon Cross [simon!=3D!moldiscovery.com]

> Hi Pavle, have you considered other methods for pKa calculation of
> drug-like molecules? We recently published a method (Milletti, F. et a= l.
> (2007) New and Original pKa Prediction Method Using Grid Molecular
> Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is used=
> in our software MoKa.

> You can see more about it at www.moldiscovery.com/soft_moka.php

> We get an RMSE of 0.4 log units for the training set of 26000 data
> points (very close to experimental error) and external prediction erro= r
> of 0.7 log units.

> Simon

> Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote:
>> Sent to CCL by: "Pavle =A0Mocilac" [pavle.mocilac2{:}mail.dcu.ie]
>> My dear colleagues,
>>
>> Recently I sent the question about the calculation of pKa of small= druglike molecules. The method asks employing high accuracy models like G1= , G2, CBS-Q in order to get accurate Free Gibbs energies using Gaussian 09,= of course..
>> Now, it seems to me that this model works OK, and fast enough for = very small molecules up to 10 or 12 atoms (hydrogen included). But, my mole= cule has 25 atoms and currently is stacked at QCISD(T)/6-31+G(d') step,= calculating something called "triples" for more than 2 days!!!!!=
>> I would like to emphasize that this job is being calculating with = Gaussian 09 (Linux version), using lots of memory () and 8 processors at SG= I Altix ICE 8200EX.
>> What is the limit of CBS-Q method regarding the number of atoms? I= s 25 atoms too much? Is there any option to use some kind of additional key= words to exclude "triples" calculation at QCISD(T)/6-31+G(d')= step? Will that exclusion hamper accurate result? Is it better to use CBS-= 4M?>
>>
>>



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY..ccl.n= et or use:
=A0 =A0 =A0http://www.ccl.net/cgi-bin/ccl/send_ccl_message

E-mail to administrators: CHEM= ISTRY-REQUEST..ccl.net or use
=A0 =A0 =A0http://www.ccl.net/cgi-bin/ccl/send_ccl_message

Subscribe/Unsubscribe:
=A0 =A0 =A0http://www.ccl.net/chemistry/sub_unsub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.n= et/jobs
Conferences: http://server.ccl.net/chemistry/announcements/co= nferences/

Search Messages: http://www.ccl.net/chemistry/searchccl/index.shtml
=A0 =A0 =A0h= ttp://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/instructions/



--0016e649d942099b62046e4682e3-- From owner-chemistry@ccl.net Thu Jul 9 12:13:01 2009 From: "Anne Rutland ccdc-announce..ccdc.cam.ac.uk" To: CCL Subject: CCL: Release of GOLD v4.1 Message-Id: <-39725-090709105626-25340-3Gv7iPWagFOS4xQ0m0WzHQ###server.ccl.net> X-Original-From: "Anne Rutland" Date: Thu, 9 Jul 2009 10:56:22 -0400 Sent to CCL by: "Anne Rutland" [ccdc-announce#ccdc.cam.ac.uk] The Cambridge Crystallographic Data Centre is pleased to announce the recent release of GOLD version 4.1 software for ligand-protein docking. GOLD v.4.1 is supplied as part of the GOLD suite, which also includes the Hermes 1.3 protein visualiser and the post-docking analysis package GoldMine 1.2. New Features included in this version of the GOLD Suite: * A new scoring function, PLP (Piecewise Linear Potential) has been added (note that this is a beta feature). Further details about this scoring function can be found in the following publication O. Korb, T. Sttzle and T. E. Exner, J. Chem. Inf. Model., 49, 84-96, 2009. * Ring conformation is now treated using a ring template library derived from the Cambridge Structural Database (CSD). If a match is found to any ligand rings, library conformations are used during the docking. It is possible for users to create their own ring template libraries. * GOLD configuration files can be set-up so that docking solutions are automatically rescored with a second scoring function. * A number of target-type specific docking(/rescoring) protocols are available from a template library. * A per-atom score output option has been added so that individual atom contributions to the scoring function can be inspected. * Interaction motifs can now be defined to bias the docking towards solutions that form particular protein-ligand binding motifs. * Flexible side-chain handling and toggling of active site waters are now available for all scoring functions. * Regression tools are included in GoldMine 1.2 that allow target specific scoring functions; discrimination models for virtual screening; and 3D QSAR models, to be created from descriptors and scoring function components calculated from the docked poses. * ROC curve graphing and calculation of enrichment metrics (AUC ROC, EF, BEDROC, MCC) are now available in GoldMine 1.2 * Interaction hot-spots derived from ligand overlay can be calculated and displayed in GoldMine 1.2. For further information please visit http://www.ccdc.cam.ac.uk/products/gold_suite. For a 60 day evaluation copy, please contact admin_+_ccdc.cam.ac.uk Best regards Anne Rutland Cambridge Crystallographic Data Centre 12 Union Rd, Cambridge CB2 1EZ From owner-chemistry@ccl.net Thu Jul 9 12:55:01 2009 From: "rocky walden rocky.walden19*o*gmail.com" To: CCL Subject: CCL: Needed free or trial version of any QSAR software Message-Id: <-39726-090708031552-6070-ohYAECAtfznXIuhX6BOBhw^server.ccl.net> X-Original-From: rocky walden Content-Type: multipart/alternative; boundary=00163645804cc02fca046e2c81d7 Date: Wed, 8 Jul 2009 12:45:41 +0530 MIME-Version: 1.0 Sent to CCL by: rocky walden [rocky.walden19]~[gmail.com] --00163645804cc02fca046e2c81d7 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Hi all, I am working on Non-selective PDE inhibitors(Xanthine Derivatives) i have synthesized a compound form caffeine family i need to find a pharmacophore for it and also study its QSAR properties I need any trial version of computation tools that are avalible for me to study them Thanks Rocky --00163645804cc02fca046e2c81d7 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit
Hi all,

I am working on Non-selective PDE inhibitors(Xanthine Derivatives) i have synthesized a compound form caffeine family i need to find a pharmacophore for it and
also study its QSAR properties
I need any trial version of computation tools that are avalible for me to study them

Thanks
Rocky

--00163645804cc02fca046e2c81d7-- From owner-chemistry@ccl.net Thu Jul 9 15:04:01 2009 From: "Tim Pyrkov pyrkov{:}nmr.ru" To: CCL Subject: CCL: Hydrophobicity Message-Id: <-39727-090709143610-1446-XnHnDmaKnB/kXR+85E3e3w^-^server.ccl.net> X-Original-From: Tim Pyrkov Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=ISO-8859-1; format=flowed Date: Thu, 09 Jul 2009 21:43:02 +0400 MIME-Version: 1.0 Sent to CCL by: Tim Pyrkov [pyrkov||nmr.ru] Dear Babak Khalili, If you need just to claculate the sum hydrophobicity of a peptide you can use one of the web-services, for example this: http://www.vcclab.org/lab/alogps/. In case you need more detailed representation you can use some molecular visualizing soft (like Sybyl) which incorporate tools for mapping molecular hydrophobicity onto its surface. In our lab we have much work with a-helical peptides and found it useful to project their hydrophobic/hydrophilic properties onto cylindrical map. Just recently we made this tool available on the internet. You can try it here: http://model.nmr.ru/platinum/. Best regards, Tim Pyrkov, Laboratory of Molecular Modeling Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry Russian Academy of Sciences From owner-chemistry@ccl.net Thu Jul 9 16:40:00 2009 From: "Babak Khalili khalili.babak^^gmail.com" To: CCL Subject: CCL: Hydrophobicity Message-Id: <-39728-090709161202-4477-zCDan1xZbpqy/AMomwo+pA..server.ccl.net> X-Original-From: Babak Khalili Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=ISO-8859-1 Date: Fri, 10 Jul 2009 00:33:52 +0430 MIME-Version: 1.0 Sent to CCL by: Babak Khalili [khalili.babak[A]gmail.com] Dear Dr. Tim Pyrkov Thank you in advance for this great kind, I hope we can work much more in near future in this area. Best regards, Babak Khalili On 7/9/09, Tim Pyrkov pyrkov{:}nmr.ru wrote: > > Sent to CCL by: Tim Pyrkov [pyrkov||nmr.ru] > Dear Babak Khalili, > > If you need just to claculate the sum hydrophobicity of a peptide you > can use one of the web-services, for example this: > http://www.vcclab.org/lab/alogps/. > > In case you need more detailed representation you can use some molecular > visualizing soft (like Sybyl) which incorporate tools for mapping > molecular hydrophobicity onto its surface. > > In our lab we have much work with a-helical peptides and found it useful > to project their hydrophobic/hydrophilic properties onto cylindrical > map. Just recently we made this tool available on the internet. You can > try it here: http://model.nmr.ru/platinum/. > > Best regards, > > Tim Pyrkov, > Laboratory of Molecular Modeling > Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry > Russian Academy of Sciences> > > From owner-chemistry@ccl.net Thu Jul 9 17:15:00 2009 From: "willsd|a|appstate.edu" To: CCL Subject: CCL:G: CBS-Q method in druglike molecules Message-Id: <-39729-090709163915-8084-pzYVOMTxrbtVkkD1ukl2+g:server.ccl.net> X-Original-From: Content-Disposition: inline Content-Language: en Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=us-ascii Date: Thu, 09 Jul 2009 19:54:58 GMT MIME-Version: 1.0 Sent to CCL by: [willsd*appstate.edu] Here is my $0.02... You would like to apply a method based on CBS-Q energies to estimate pKa values for molecules with as many as 25 atoms, and the QCISD(T)/6-31+G(d) step is taking a long time. You have G09 and would like to do this calculation with this software, so as to come as close as possible to reproducing the methods used for similar calculations in a reference paper... You have not mentioned this, but you may run into disk limitation problems... this is quite likely with a 32 bit version on G09, since the 32 bit version can address only a limited amount of disk, and QCISD(T) will need a lot of disk. (I should mention that I have not used G09... but previous versions of Gxx had this limitation on the 32 bit versions, and I'd guess that 32 bit G09 does too.) With the 64 bit version, running on a 64 bit machine with a 64 bit OS, you will run out of physical disk long before you run out address space. So one repliers comments about 32 bit vs 64 bit is quite pertinent to this calculation. The CBS methods (and other high accuracy combined methods (G1, G2, G3, multi-coefficient G3, ....) all try, by various methods, to get the energy accuracy of a very large basis set combined with a very accurate energy method, without actually doing such a calculation, since the required computing resources exceed what anyone has. All such methods depend on various post-HF energy calculations, and these depend on computing the effects on the energy of substituting electrons from occupied (in the underlying HF wavefunction) molecular orbitals into higher energy, unoccupied orbitals. The SD in QCISD(T) means that all possible subsitutions of single electrons, and pairs of electrons are included in a self-consistent manner, and the (T) means that once the SD part has converged, the effect of substituting three electrons (hence triples) is estimated with perturbation theory. The cost of such a calculation is roughly proportional to the sixth power of the number of basis functions (this is included in the G09 output), and hence proportional to the sixth power of the number of atoms. The CBS method (and G1, G2....) has other steps as well, and this will have other scaling behavior, so the overall scaling for the entire CBS-Q method is not easy to predict. One replyer suggested replacing the QCISD(T) step with QCISD... this can be made to give a possibly useful method, but it will not be CBS-Q, and will not be comparable to your reference paper. I'd suggest trying a scaled down version of your drug-like molecule (replace methyl or ethyl groups with H, etc), and see if you can get the atom count down to 12 or so. Do the CBS-Q calculation on this version, note the time required, add a methyl or ethyl back in, repeat the CBS-Q and timing. You can then estimate the scaling, and use this to estimate how long your full 25 atom molecule might take. My rule of thumb is that (except in unusual circumstances) if it is going to take longer than a week, I will have forgotten what I wanted to calculate before it is done, and I am not interested. For some quite clear discussion on the calculations that lie inside combined methods like CBS-Q, Frank Jensen's relatively new book (Introduction to Computational Chemistry) is pretty good. (IMHO, it is an introduction in the sense that he introduces nearly all of the relevant equations, not in the sense of simplified. I appreciate his approach a great deal.) Good luck on your hunt for pKa estimates... Steve Williams ----- Original Message ----- > From: "Pavle Mocilac pavle.mocilac2^-^mail.dcu.ie" Date: Thursday, July 9, 2009 10:03 am Subject: CCL:G: CBS-Q method in druglike molecules To: "Williams, Steve " > > Sent to CCL by: Pavle Mocilac [pavle.mocilac2===mail.dcu.ie] > Dear Simon, > > Thank you for the information about this software, but no, I have > not and will not consider any other methods beside those that can > be made with Gaussian 03 or 09 and described in "CaballeroNA, > Melendez FJ, Munoz-Cara C, et al. Biophys. Chem 124 (2) > 155-160". > We do not have any intention to buy any new additional software but > to use the one we have. Is seems to me that there are numerous > people on this CCL that behave as salesman and always want to > solve my problem by selling > me some additional software. > The reason we are using Gaussian is deeper, and for the sake of > consistency and comparability we do not have attention to use so > many different programmes that works in different way. > I would like to repeat that CBS-Q is my main problem and therefore > I will be very happy and very grateful if there is somebody in this > CCL who can help me with CBS-Q calculations in Gaussian. > > > Best regards, > > Pavle Mocilac > > ============================================ > Pavle Mocilac > Postgraduate Researcher > T3 - Targeted Therapeutics and Theranostics > Room X249, School of Chemistry > Dublin City University > Dublin 9, Dublin, Ireland > mobile: +353872167022 > mailto:pavle.mocilac2~!~mail.dcu.ie > ============================================ > Wednesday, July 8, 2009, 7:21:16 PM, you wrote: > > > > Sent to CCL by: Simon Cross [simon!=!moldiscovery.com] > > > Hi Pavle, have you considered other methods for pKa calculation > of > > drug-like molecules? We recently published a method (Milletti, F. > et al. > > (2007) New and Original pKa Prediction Method Using Grid > Molecular > > Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is > used> in our software MoKa. > > > You can see more about it at www.moldiscovery.com/soft_moka.php > > > We get an RMSE of 0.4 log units for the training set of 26000 > data > > points (very close to experimental error) and external prediction > error> of 0.7 log units. > > > Simon > > > Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote: > >> Sent to CCL by: "Pavle Mocilac" [pavle.mocilac2{:}mail.dcu.ie] > >> My dear colleagues, > >> > >> Recently I sent the question about the calculation of pKa of > small druglike molecules. The method asks employing high accuracy > models like G1, G2, CBS-Q in order to get accurate Free Gibbs > energies using Gaussian 09, of course.. > >> Now, it seems to me that this model works OK, and fast enough > for very small molecules up to 10 or 12 atoms (hydrogen included). > But, my molecule has 25 atoms and currently is stacked at > QCISD(T)/6-31+G(d') step, calculating something called "triples" > for more than 2 days!!!!! > >> I would like to emphasize that this job is being calculating > with Gaussian 09 (Linux version), using lots of memory () and 8 > processors at SGI Altix ICE 8200EX. > >> What is the limit of CBS-Q method regarding the number of atoms? > Is 25 atoms too much? Is there any option to use some kind of > additional keywords to exclude "triples" calculation at QCISD(T)/6- > 31+G(d') step? Will that exclusion hamper accurate result? Is it > better to use CBS-4M?> > >> > >> > > > > -= This is automatically added to each message by the mailing > script =- > To recover the email address of the author of the message, please > changethe strange characters on the top line to the ### sign. You can > also> Conferences: > http://server.ccl.net/chemistry/announcements/conferences/> > > From owner-chemistry@ccl.net Thu Jul 9 19:06:00 2009 From: "Kefa Lu kefa.lu!A!gmail.com" To: CCL Subject: CCL: pKa calculation of strong acid Message-Id: <-39730-090709190418-8572-kDgDa0HPY2kbOaBmMA3mqw*server.ccl.net> X-Original-From: Kefa Lu Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=ISO-8859-1; format=flowed Date: Thu, 09 Jul 2009 15:52:45 -0600 MIME-Version: 1.0 Sent to CCL by: Kefa Lu [kefa.lu!^!gmail.com] Hi, all, Really appreciate for all the replys. What I actually need to know is the difference of dissociation energy H+ or binding energy of H+ to Cl- in water and in a protein I'm studying right now. A problem is, HCl usually completely dissociates into H+ and Cl-. So how can I find a way to estimate the free energy of binding a H+ to Cl-? Just calculate the potential of mean force using umbrella sampling method? Or is there any better idea to do this? Thanks. Kefa VITORGE Pierre 094605 Pierre.VITORGE%a%cea.fr wrote: > Sent to CCL by: "VITORGE Pierre 094605" [Pierre.VITORGE*cea.fr] > Dave, > > pKa = -6 in aqueous solution, means it cannot be measured. We are both right: the value cannot be measured, but it might very well be the good value, which I am explaining below. > > pKa = -lgKa is a definition of pKa, where > Ka = [H+] [Cl-] / [HCl] > [X] is the concentration of species X, here X is an aqueous specie (H+, Cl- or HCl). For simplicity I am using concentrations meaning the (small) activity corrections are included in Ka. > > A first (chemical) meaning of pKa is (pH) half point reaction, namely for 50% dissociation [Cl-]_1/2 = [HCl]_1/2, where subscript _1/2 is to stress concentrations are for special (chemical) conditions. Consequently > Ka = [H+]_1/2 equivalently > pKa = -lg[H+]_1/2 and > pKa° = pH_1/2 a well known formula in solution chemistry, where superscript° is for standard conditions (infinite dilution where -lg[H+] -> pH) > > In this respect pKa of strong acid has no finite value, since strong acid means completely dissociated in any chemical conditions, namely [HCl]=0; which is indeed not possible for finite values of Ka. > However, Ka can directly be measured only when the dissociated form can be detected, typically if the detection limit is 1% only pKa > -2 can be directly measured, and "no finite value" actually means "> -2". > > (Partial) conclusion: do not write pKa of strong acid; it is chocking for those who currently measure or use pKas knowing the meanings of the concept. > > Another meaning of pKa is form thermodynamics: delta_rG = -RTlnK here applied to K = Ka. You can very well calculate > delta_rGa = G(H+) + G(Cl-) - G(HCl) > (again note they are aqueous species, especially not HCl(g)), where G(X) can be taken from thermochemical data bases (as G(X) = delta_fG(X)) or calculated by molecular modelling. > Now imagine delta_rGa have been estimated in this way giving pKa=-6. This means that in the most favourable conditions to form HCl, typically 10 mol.L-1 total HCl concentration, the concentrations of the actual species in solution would be > [H+] # 10 mol.L-1 > [Cl-] # 10 mol.L-1 > [HCl] = 10^-6 mol.L-1, which can probably not be detected, but might very well exist. The reasoning still stand even for concentrations less than Avogadro Number per litre, and the result is meaningful: the (Ka) reaction can be used in thermodynamic cycles, typically with another acid-base reaction to decide in which way the proton will be transferred. > > Conclusion: what is your actual problem? Do you really need the pKa of HCl in liquid water? Or is it only part of your calculations? In this later case can it be cancelled out by using other calculation strategies? > This was actually the origin of my remark. > > Pierre Vitorge > > -----Message d'origine----- > De : owner-chemistry+pierre.vitorge==cea.fr*|*ccl.net [mailto:owner-chemistry+pierre.vitorge==cea.fr*|*ccl.net] De la part de case case~!~biomaps.rutgers.edu > Envoyé : mercredi 8 juillet 2009 16:44 > À : VITORGE Pierre 094605 > Objet : CCL: pKa calculation of strong acid > > > Sent to CCL by: case [case##biomaps.rutgers.edu] > > On Tue, Jul 07, 2009, VITORGE Pierre 094605 Pierre.VITORGE^^cea.fr wrote: > >> The pKa of a strong acid have no finite value, namely it does not exist. >> > > These values certainly exist. For the HCl example that was in the > original post, the pKa in aqueous solution is around -6, although this > seems to have a big uncertainty. You would have to spend some time to > find the original determination -- a secondary reference to start from > would be Albert and Serjeant, "The Determination of Ionization Constants", > Chapman & Hall, 1984. > > ....dave casehttp://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/spammers.txt > > > -=his is automatically added to each message by the mailing script =-> > > From owner-chemistry@ccl.net Thu Jul 9 20:11:01 2009 From: "Pavle Mocilac pavle.mocilac2**mail.dcu.ie" To: CCL Subject: CCL:G: CBS-Q pka calculation Message-Id: <-39731-090709194557-26826-3bMyOfKH7ok6n7Bjyh5K8g[]server.ccl.net> X-Original-From: Pavle Mocilac Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=windows-1252 Date: Fri, 10 Jul 2009 00:25:26 +0100 MIME-Version: 1.0 Sent to CCL by: Pavle Mocilac [pavle.mocilac2:_:mail.dcu.ie] Dear James I'm using 64 bit machine and 64 bit UNIX operating system (Silicon Graphics Altix ICE 8200XE). I can not improve my hardware, and I'm quite sure that the better hardware I can not get. I'm using Gaussian 09. This is the specification in gjf file: %chk=2AP_W_CBSQ.chk %mem=256MW %nproc=8 # cbs-q scrf=(solvent=water,smd) geom=connectivity The structures are previously optimized using DFT B3LYP/6-311++G level, without frequency calculations. I think it is nicely high level of calculation itself. Best regards, Pavle Mocilac ============================================ Pavle Mocilac Postgraduate Researcher T3 - Targeted Therapeutics and Theranostics Room X249, School of Chemistry Dublin City University Dublin 9, Dublin, Ireland mobile: +353872167022 mailto:pavle.mocilac2-#-mail.dcu.ie ============================================ Thursday, July 9, 2009, 3:49:03 PM, you wrote: > Sent to CCL by: "James Robinson" [james.robinson_-_prosonix.co.uk] > Dear Pavle, > Your problem is related to hardware, i suspect. Remember that there > may be options within gaussian to make use of more efficient memory, > are you optmising the starting structures with a lower level of > theory, are you using options like symmetry, are your calculations > running as efficiently as possible? Would optimising strutures with > lower level of theory help in getting the strutures for the > calculation in better time. Are using ramen options in freq > calculations. If you cannot improve the efficiency of your > calculation, how can you improve your hardware, are you using 32 bit > software on a 64 bit capable machine.. > Regards > James > Prosonix > Oxoford.> Conferences: > http://server.ccl.net/chemistry/announcements/conferences/ From owner-chemistry@ccl.net Thu Jul 9 20:45:01 2009 From: "Pavle Mocilac pavle.mocilac2[]mail.dcu.ie" To: CCL Subject: CCL:G: CBS-Q method in druglike molecules Message-Id: <-39732-090709194552-26768-/JcyxNyDOyNqXf9Ij34SvQ^-^server.ccl.net> X-Original-From: Pavle Mocilac Content-Transfer-Encoding: 8bit Content-Type: text/html; charset=iso-8859-1 Date: Thu, 9 Jul 2009 23:57:20 +0100 MIME-Version: 1.0 Sent to CCL by: Pavle Mocilac [pavle.mocilac2%mail.dcu.ie] Re: CCL:G: CBS-Q method in druglike molecules

Dear Gonçalo 


Thank you for the information, but I do not know how to specify exclusion of triples in QCISD(T) calculation since it is a part of CBS-Q calculation. I tried at Gaussian web page to find something useful but I found nothing... 



Best regards,


Pavle Mocilac


Thursday, July 9, 2009, 3:16:55 PM, you wrote:

dear pavle,


as far as i'm aware (someone correct me) the triples corection you refer to come from the '(T)' part in the QCISD(T) keyword. Using QCISD only (no T into it) would not require triples correction. can you afford to use the QCISD instead of the QCISD(T) ? I know this is the simple answer, but it's the only one I can come up with.


the triples come from J. Chem. Phys., 87 (1987) 5968-75..


hope it helps at all,

gonçalo






2009/7/9 Pavle Mocilac pavle.mocilac2^-^mail.dcu.ie <owner-chemistry : ccl.net>



Sent to CCL by: Pavle Mocilac [pavle.mocilac2===mail.dcu.ie]

Dear Simon,


Thank you for the information about this software, but no, I have not and will not consider any other methods beside those that can be made with Gaussian 03 or 09 and described in "CaballeroNA, Melendez FJ, Munoz-Cara C, et al.   Biophys. Chem  124  (2)

155-160".

We do not have any intention to buy any new additional software but to use the one we have. Is seems to me that there are numerous people on this CCL  that behave as salesman and always want to solve my problem by selling

me some additional software.

The reason we are using Gaussian is deeper, and for the sake of consistency and comparability we do not have attention to use so many different programmes that works in different way.

I would like to repeat that CBS-Q is my main problem and therefore I will be very happy and very grateful if there is somebody in this CCL who can help me with CBS-Q calculations in Gaussian.



Best regards,


Pavle Mocilac


============================================

Pavle Mocilac

Postgraduate Researcher

T3 - Targeted Therapeutics and Theranostics

Room X249, School of Chemistry

Dublin City University

Dublin 9, Dublin, Ireland

mobile: +353872167022

mailto:pavle.mocilac2~!~mail.dcu.ie

============================================

Wednesday, July 8, 2009, 7:21:16 PM, you wrote:



> Sent to CCL by: Simon Cross [simon!=!moldiscovery.com]


> Hi Pavle, have you considered other methods for pKa calculation of

> drug-like molecules? We recently published a method (Milletti, F. et al.

> (2007) New and Original pKa Prediction Method Using Grid Molecular

> Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is used

> in our software MoKa.


> You can see more about it at www.moldiscovery.com/soft_moka.php


> We get an RMSE of 0.4 log units for the training set of 26000 data

> points (very close to experimental error) and external prediction error

> of 0.7 log units.


> Simon


> Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote:

>> Sent to CCL by: "Pavle  Mocilac" [pavle.mocilac2{:}mail.dcu.ie]

>> My dear colleagues,

>>

>> Recently I sent the question about the calculation of pKa of small druglike molecules. The method asks employing high accuracy models like G1, G2, CBS-Q in order to get accurate Free Gibbs energies using Gaussian 09, of course..

>> Now, it seems to me that this model works OK, and fast enough for very small molecules up to 10 or 12 atoms (hydrogen included). But, my molecule has 25 atoms and currently is stacked at QCISD(T)/6-31+G(d') step, calculating something called "triples" for more than 2 days!!!!!

>> I would like to emphasize that this job is being calculating with Gaussian 09 (Linux version), using lots of memory () and 8 processors at SGI Altix ICE 8200EX.

>> What is the limit of CBS-Q method regarding the number of atoms? Is 25 atoms too much? Is there any option to use some kind of additional keywords to exclude "triples" calculation at QCISD(T)/6-31+G(d') step? Will that exclusion hamper accurate result? Is it better to use CBS-4M?>

>>

>>




E-mail to subscribers: CHEMISTRY : ccl.net or use:

     http://www.ccl.net/cgi-bin/ccl/send_ccl_message


E-mail to administrators: CHEMISTRY-REQUEST : ccl.net or use

     http://www.ccl.net/cgi-bin/ccl/send_ccl_message


Subscribe/Unsubscribe:

     http://www.ccl.net/chemistry/sub_unsub.shtml


Before posting, check wait time at: http://www.ccl.net


Job: http://www.ccl.net/jobs

Conferences: http://server.ccl.net/chemistry/announcements/conferences/


Search Messages: http://www.ccl.net/chemistry/searchccl/index.shtml

     http://www.ccl.net/spammers.txt


RTFI: http://www.ccl.net/chemistry/aboutccl/instructions/
From owner-chemistry@ccl.net Thu Jul 9 21:21:01 2009 From: "Pavle Mocilac pavle.mocilac2{}mail.dcu.ie" To: CCL Subject: CCL:G: CBS-Q method in druglike molecules Message-Id: <-39733-090709194547-26683-8+TsXXAxM91YzPNeHCEPXQ[#]server.ccl.net> X-Original-From: Pavle Mocilac Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=windows-1252 Date: Fri, 10 Jul 2009 00:45:15 +0100 MIME-Version: 1.0 Sent to CCL by: Pavle Mocilac [pavle.mocilac2()mail.dcu.ie] Dear Steve, Thank you for this nice mail, you give me some new useful information. As I mentioned im my previous mail, I'm using 64 bit machine and 64 bit UNIX operating system (Silicon Graphics Altix ICE 8200XE). I can not improve my hardware, and I'm quite sure that the better hardware I can not get. The memory I'm allocating is 256MW and I am using 8 processors - it is a maximum regarding the number of processor power. Therefore, I'm using quite awesome hardware/software performances. Do you think that QCISD(T) method will be possible to replace with QCISD method? How to specify T exclusion in QCISD(T) step of CBS-Q protocol using keywords? Regarding the scaling the molecule, I was developing a model using the same molecules described in reference paper, aminopyridines (13 atoms) and for them, the calculation lasts up to 45 minutes maximum, including in a solvent conditions. But my drug-like molecules contain 25 atoms, but in that case, calculation goes to more than 3 days. In fact, as I have limited time for one job I have to say that I never finished. In the paper I'm using as a reference G1 level was used and they got nice correlation (pKa{calc} vs pKa{exp}). My idea, using CBS-Q, was, in fact, to obtain even better model... Maybe to use CBS-4M or to try G1? Although mistakes are supposed to be bigger... Best regards, Pavle Mocilac ============================================ Pavle Mocilac Postgraduate Researcher T3 - Targeted Therapeutics and Theranostics Room X249, School of Chemistry Dublin City University Dublin 9, Dublin, Ireland mobile: +353872167022 mailto:pavle.mocilac2(-)mail.dcu.ie ============================================ Thursday, July 9, 2009, 8:54:58 PM, you wrote: > Sent to CCL by: [willsd*appstate.edu] > Here is my $0.02... > You would like to apply a method based on CBS-Q energies to estimate pKa values > for molecules with as many as 25 atoms, and the QCISD(T)/6-31+G(d) step is > taking a long time. You have G09 and would like to do this calculation with > this software, so as to come as close as possible to reproducing the methods > used for similar calculations in a reference paper... > You have not mentioned this, but you may run into disk limitation problems... > this is quite likely with a 32 bit version on G09, since the 32 bit version can > address only a limited amount of disk, and QCISD(T) will need a lot of disk. (I > should mention that I have not used G09... but previous versions of Gxx had this > limitation on the 32 bit versions, and I'd guess that 32 bit G09 does too.) With > the 64 bit version, running on a 64 bit machine with a 64 bit OS, you will run > out of physical disk long before you run out address space. So one repliers > comments about 32 bit vs 64 bit is quite pertinent to this calculation. > The CBS methods (and other high accuracy combined methods (G1, G2, G3, > multi-coefficient G3, ....) all try, by various methods, to get the energy > accuracy of a very large basis set combined with a very accurate energy method, > without actually doing such a calculation, since the required computing > resources exceed what anyone has. All such methods depend on various post-HF > energy calculations, and these depend on computing the effects on the energy of > substituting electrons from occupied (in the underlying HF wavefunction) > molecular orbitals into higher energy, unoccupied orbitals. The SD in QCISD(T) > means that all possible subsitutions of single electrons, and pairs of electrons > are included in a self-consistent manner, and the (T) means that once the SD > part has converged, the effect of substituting three electrons (hence triples) > is estimated with perturbation theory. The cost of such a calculation is > roughly proportional to the sixth power of the number of basis functions (this > is included in the G09 output), and hence proportional to the sixth power of the > number of atoms. The CBS method (and G1, G2....) has other steps as well, and > this will have other scaling behavior, so the overall scaling for the entire > CBS-Q method is not easy to predict. > One replyer suggested replacing the QCISD(T) step with QCISD... this can be made > to give a possibly useful method, but it will not be CBS-Q, and will not be > comparable to your reference paper. > I'd suggest trying a scaled down version of your drug-like molecule (replace > methyl or ethyl groups with H, etc), and see if you can get the atom count down > to 12 or so. Do the CBS-Q calculation on this version, note the time required, > add a methyl or ethyl back in, repeat the CBS-Q and timing. You can then > estimate the scaling, and use this to estimate how long your full 25 atom > molecule might take. My rule of thumb is that (except in unusual circumstances) > if it is going to take longer than a week, I will have forgotten what I wanted > to calculate before it is done, and I am not interested. > For some quite clear discussion on the calculations that lie inside combined > methods like CBS-Q, Frank Jensen's relatively new book (Introduction to > Computational Chemistry) is pretty good. (IMHO, it is an introduction in the > sense that he introduces nearly all of the relevant equations, not in the sense > of simplified. I appreciate his approach a great deal.) > Good luck on your hunt for pKa estimates... > Steve Williams > ----- Original Message ----- >> From: "Pavle Mocilac pavle.mocilac2^-^mail.dcu.ie" > Date: Thursday, July 9, 2009 10:03 am > Subject: CCL:G: CBS-Q method in druglike molecules > To: "Williams, Steve " >> >> Sent to CCL by: Pavle Mocilac [pavle.mocilac2===mail.dcu.ie] >> Dear Simon, >> >> Thank you for the information about this software, but no, I have >> not and will not consider any other methods beside those that can >> be made with Gaussian 03 or 09 and described in "CaballeroNA, >> Melendez FJ, Munoz-Cara C, et al. Biophys. Chem 124 (2) >> 155-160". >> We do not have any intention to buy any new additional software but >> to use the one we have. Is seems to me that there are numerous >> people on this CCL that behave as salesman and always want to >> solve my problem by selling >> me some additional software. >> The reason we are using Gaussian is deeper, and for the sake of >> consistency and comparability we do not have attention to use so >> many different programmes that works in different way. >> I would like to repeat that CBS-Q is my main problem and therefore >> I will be very happy and very grateful if there is somebody in this >> CCL who can help me with CBS-Q calculations in Gaussian. >> >> >> Best regards, >> >> Pavle Mocilac >> >> ============================================ >> Pavle Mocilac >> Postgraduate Researcher >> T3 - Targeted Therapeutics and Theranostics >> Room X249, School of Chemistry >> Dublin City University >> Dublin 9, Dublin, Ireland >> mobile: +353872167022 >> mailto:pavle.mocilac2~!~mail.dcu.ie >> ============================================ >> Wednesday, July 8, 2009, 7:21:16 PM, you wrote: >> >> >> > Sent to CCL by: Simon Cross [simon!=!moldiscovery.com] >> >> > Hi Pavle, have you considered other methods for pKa calculation >> of >> > drug-like molecules? We recently published a method (Milletti, F. >> et al. >> > (2007) New and Original pKa Prediction Method Using Grid >> Molecular >> > Interaction Fields. J. Chem. Inf. Model., 47, 2172-2181) which is >> used> in our software MoKa. >> >> > You can see more about it at www.moldiscovery.com/soft_moka.php >> >> > We get an RMSE of 0.4 log units for the training set of 26000 >> data >> > points (very close to experimental error) and external prediction >> error> of 0.7 log units. >> >> > Simon >> >> > Pavle Mocilac pavle.mocilac2]|[mail.dcu.ie wrote: >> >> Sent to CCL by: "Pavle Mocilac" [pavle.mocilac2{:}mail.dcu.ie] >> >> My dear colleagues, >> >> >> >> Recently I sent the question about the calculation of pKa of >> small druglike molecules. The method asks employing high accuracy >> models like G1, G2, CBS-Q in order to get accurate Free Gibbs >> energies using Gaussian 09, of course.. >> >> Now, it seems to me that this model works OK, and fast enough >> for very small molecules up to 10 or 12 atoms (hydrogen included). >> But, my molecule has 25 atoms and currently is stacked at >> QCISD(T)/6-31+G(d') step, calculating something called "triples" >> for more than 2 days!!!!! >> >> I would like to emphasize that this job is being calculating >> with Gaussian 09 (Linux version), using lots of memory () and 8 >> processors at SGI Altix ICE 8200EX. >> >> What is the limit of CBS-Q method regarding the number of atoms? >> Is 25 atoms too much? Is there any option to use some kind of >> additional keywords to exclude "triples" calculation at QCISD(T)/6- >> 31+G(d') step? Will that exclusion hamper accurate result? Is it >> better to use CBS-4M?> >> >> >> >> >> >> >> >> -= This is automatically added to each message by the mailing >> script =- >> To recover the email address of the author of the message, please >> changethe strange characters on the top line to the _._ sign. You can >> also> Conferences: >> http://server.ccl.net/chemistry/announcements/conferences/> Conferences: > http://server.ccl.net/chemistry/announcements/conferences/