From owner-chemistry@ccl.net Wed Oct 21 03:03:01 2009 From: "Andreas Klamt klamt++cosmologic.de" To: CCL Subject: CCL: solvation energy in mopac Message-Id: <-40507-091021025710-11807-w1Mf3llePDy3lQovPh2Q+A%server.ccl.net> X-Original-From: Andreas Klamt Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-15; format=flowed Date: Wed, 21 Oct 2009 08:18:57 +0200 MIME-Version: 1.0 Sent to CCL by: Andreas Klamt [klamt#%#cosmologic.de] Dear Daniel, as author of COSMO in MOPAC I like to emphazise that the COSMO in MOPAC=20 is not parameterized for the quantitative calculation of dG_solv. It is=20 definitely good for getting the calculations closer to solvation than=20 just a gasphase calculation, but for two reasons I recommend higher=20 levels for optimal calculation of dG_solv: 1) The electrostatics of semi-empirical methods is usually not very=20 accurate (because this mostly is not in the target properties of the=20 parameterizations). At least in AM1, PM3, and PM5 (we did not test PM6)=20 dipole moments of some functional groups can be off by one Debye (e.g.=20 nitro groups) and if the dipole moment is of that much the solvation=20 energy will be very bad. This is the reason why I left the=20 semi-empirical level for quantitative solvation calculations a few years = after developing COSMO in MOPAC. And most likely this is also why the=20 Minnesota group developed parameterized charge models based on=20 semi-empirical methods in order to improve the SMx models. Since HF is=20 generally overestimating polarities, this is also not extremely=20 reliable. The good message is that DFT is very reliable in that regard,=20 and this quite independent of the special functional. From my=20 perspective DFT is currently the method of choice, although for getting=20 the solvations energies much better than 0.3 kcal/mol one will need even = more accurate methods than DFT. By the way, a semi-emirical geometry=20 furnished with a single point DFT calculation can be a very good choice=20 for calculating solvation energies. 2) The second reason is that a dielectric continuum solvation model=20 alone cannot give the complete solvation free energies. One needs in=20 addition "non-electrostatic contributions". While in the SMx and PCM=20 models these have been parameterized with many additional, solute and=20 solvent specific parameters and descriptors, I have never implemented=20 such non-electrostatic terms in MOPAC, at least not quantitatively.=20 MOPAC/COSMO thus does not have these terms. Instead, I have attacked=20 this question with the much more rigorous COSMO-RS approach, which=20 combines COSMO with a statistical thermodynamics of interacting surfaces = (using the COSMO polarization charges for the quantification of the=20 interactions) and thus describes solutes and solvents on the same=20 quantum-chemical footing, gives enthalpies and entropies, temperature=20 dependencies, and allows for the treatment of mixtures. But due to the=20 earlier finding of 1) I have only developed COSMO-RS quantitatively on=20 the DFT level. Based on a comparison on ~ 2500 dG_solv predictions (see=20 "On the Performance of Continuum Solvation Methods. A Comment on=20 "Universal Approaches to Solvation Modeling"", Acc. Chem. Res., 2009, 42 = (4), pp 489--492) and on the outcome of the recent SAMPL blindtest (=20 publications in preparation) I dare to say that COSMO-RS currently is=20 the most accurate method for the prediction of dG_solv. Andreas Daniel Glossman-Mitnik dglossman%a%gmail.com schrieb: > Dear netters: > > How should one calculate the solvation energy of a given molecule=20 > using MOPAC 2009? > It is enough with doing one calculation in gas phase and another in=20 > solvent and then > substracting both heats of formation? > Any example input related to these questions will be appreciated. > > Best regards, > > Daniel > > ***********************************************************************= ************************************** > Dr. Daniel Glossman-Mitnik: > Centro de Investigaci=F3n en Materiales Avanzados, SC > Grupo NANOCOSMOS - Nanotecnolog=EDa Computacional, Simulaci=F3n y Model= ado=20 > Molecular > Miguel de Cervantes 120 - Complejo Industrial Chihuahua - Chihuahua,=20 > Chih 31109, Mexico > Phone: +52 614 4391151 Secretary/FAX: +52 614 4394852 Lab: +52 614 4394= 805 > E-mail: daniel.glossman[A]cimav.edu.mx=20 > dglossman[A]gmail.com=20 > > WWW: http://www.cimav.edu.mx/cv/daniel.glossman > http://blogs.cimav.edu.mx/daniel.glossman > ***********************************************************************= ************************************** --=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 Wed Oct 21 06:20:01 2009 From: "Berger Raphael berger#,#chem.helsinki.fi" To: CCL Subject: CCL: physical interpretation of core-electron correlation-effect Message-Id: <-40508-091021052048-22937-VDZzTojmaUSjWlO/7lle6A]|[server.ccl.net> X-Original-From: Berger Raphael Content-Type: TEXT/PLAIN; format=flowed; charset=US-ASCII Date: Wed, 21 Oct 2009 11:42:34 +0300 (EEST) MIME-Version: 1.0 Sent to CCL by: Berger Raphael [berger++chem.helsinki.fi] Dear CCL-Readers, is anyone aware of o a physical interpretation of core electron correlation effects? One advantage of ab-initio methods is that they go hand in hand with the MO-picture building up the first order reference functions for the wavefunction. In this picture the van-der-Waals atttraction or Pauli-repulsion effects find their counterparts in the correlation contributions of the diffuse outer orbitals. My question is now, if there is a strong (lets say for example: structural) influence of core electron correlation found in a molecule, can that be assigned to some certain physical effect which has a name of its own. My interpretation would be, that the effect is somehow connected to a weaker effective nuclear charge shielding due to the correlation but I'm not sure ... Thank You for reading and best regards R. Berger From owner-chemistry@ccl.net Wed Oct 21 06:54:01 2009 From: "Vincent Leroux vincent.leroux++loria.fr" To: CCL Subject: CCL: Increased ligand-protein docking accuracy Message-Id: <-40509-091021050111-17026-4Tp7HZ5ilwgcBnFeeMA/3g(_)server.ccl.net> X-Original-From: Vincent Leroux Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=ISO-8859-1; format=flowed Date: Wed, 21 Oct 2009 11:00:56 +0200 MIME-Version: 1.0 Sent to CCL by: Vincent Leroux [vincent.leroux^_^loria.fr] That was too good to be true! I think the methodology section of any program evaluating the performance of a docking experiment is always the most interesting part... An excerpt from the paper (page 3): " The 53 complexes used in this study were all characterized by a resolution below 3.2 Å. The complexes were chosen partly from the AutoDock 3.0 calibration set [68], from a recently published paper examining different docking software [13] and from the core set of PDBbind Database [69]. The chosen structures possess structurally diverse ligands in complex with a heterogeneous collection of proteins (see Table 1). It should be noted that for some structures with lower resolution (although chosen from the AutoDock 3.0 calibration set), an RMSD-based comparison of docked versus experimental structure might not always lead to a meaningful result as partial occupancies might occur that are not reflected by a single ligand structure. " - A 3.2 Å crystal structure as part of a test set? That's a good start... - Why well-known curated, diverse and publicly available test sets such as the DUD or the Astex test set were not used? - If the authors recognize that structures with lower resolution might not be meaningful, why do they still use them? - If the authors know that dubious crystal structures are part of the validation set of the docking program they use, then why not comparing their results to those of other docking programs such as GOLD or Glide that were more rigorously validated? - Have the authors questioned themselves why GOLD consistently gets good results in docking benchmarking reviews (usually better than AutoDock) despite using ridiculously oversimplified scoring functions that do not even take partial charges into account? - The authors are not even aware that a RMSD-based comparison of docked vs. experimental structure is a big mistake regardless of the experimental structures quality... See for example http://dx.doi.org/10.1021/ci600342e Later in the paper the authors explain that they actually get worse results regarding the correlation between the scoring function and the experimental binding energies by using MOPAC charges, but wait! It's all the fault of the scoring function, as it has been parameterized using Gasteiger charges! A perfect example of blind faith in the starting hypothesis (docking programs WILL be more accurate with better partial charges). Everybody knows that scoring functions are the Achille's heel of docking programs, and if better partial charges or a more accurate definition of the active site (including, but not limited to, better partial charges) were miracle solutions, it would have been reported long ago... (not saying this is necessarily useless) Then the authors proceed to: if we use RMSDs to evaluate performance, it all gets better with accurate charges... To summarize: - a bad reference set is built (talk about reinventing the wheel... and failing to do so correctly...) - the methodology for evaluating docking relevance (quantitative experimental binding energy/scoring function comparison, then geometrical accuracy assessment using RMSD calculations) could not be worse - mixed results are obtained: the binding energy correlation is actually worse, and with RMSDs it is a bit better (which by all means is not a significant result) - conclusion: "...the docking accuracy in regard to complex geometry......significantly increased when partial charges of the ligands and proteins were calculated with the semi-empirical PM6 method." This is one of the weakest papers I have read recently... this kind is not so uncommon is the first issue of a new journal, and most of the time the journal does not live long. To be fair, some of the papers of this issue do seem interesting and are worth reading. For example, there is an approach for post-docking filtering aiming to avoid the classical scoring function relevance trap http://www.jcheminf.com/content/1/1/6 (I do not know if that paper is good, but sure the issue is relevant), and the description of an interesting technique for the protonation of protein-ligand complexes http://www.jcheminf.com/content/1/1/13 (I am confident in that one, as it comes from Matthias Rarey). Improving ligand charges relevance prior to docking certainly will not hurt, but for the moment, IMHO, feeding the typical docking software scoring function with MOPAC2009-computed charges is like "donner de la confiture à des cochons" as we say in France ;-) Regards, Vincent Leroux David Gallagher gallagher.da{:}gmail.com a écrit : > *Subject: Increased ligand-protein docking accuracy > > *A recent study (1) has concluded that: "...the docking accuracy in > regard to complex geometry......significantly increased when partial > charges of the ligands *and proteins* were calculated with the > semi-empirical PM6 method." > > This new methodology using MOPAC2009 > PM6 partial charges in AutoDock > for both ligand and protein, is now implemented in "DockingServer" > which is available on-line at http://www.dockingserver.com/web? > > (1) /Journal of Cheminformatics/ 2009, *1:*15, > doi:10.1186/1758-2946-1-15 > http://www.jcheminf.com/content/1/1/15 > > David Gallagher > CACheResearch.com From owner-chemistry@ccl.net Wed Oct 21 20:21:00 2009 From: "bahman roostaei bahman.roostaei[]utsa.edu" To: CCL Subject: CCL:G: CI coefficients in natural orbital representation (Gaussian Zindo) Message-Id: <-40510-091021190251-20986-vuRGkFisi0kNuf9WBM6Ngg-*-server.ccl.net> X-Original-From: "bahman roostaei" Date: Wed, 21 Oct 2009 19:02:47 -0400 Sent to CCL by: "bahman roostaei" [bahman.roostaei]^[utsa.edu] Hi everyone I am using Gaussian 03 ... I am trying to calculate excited states of a series of photochromic molecules using Zindo . I need to find the CI coefficients in natural orbital representation. I know how to find the naturals orbitals and punch them in the card and then read them in a following zindo. Is that a right way to find the CI's in NO representation ? I am not getting sensible results. anybody is familiar with this issue ? Thanks. Bahman.