From moret@far.ruu.nl Tue Jul 5 17:50:39 1994 Received: from accucx.cc.ruu.nl for moret@far.ruu.nl by www.ccl.net (8.6.9/930601.1506) id RAA19658; Tue, 5 Jul 1994 17:35:33 -0400 Received: from far.ruu.nl (ruucmc.far.ruu.nl) by accucx.cc.ruu.nl with SMTP id AA25807 (5.65c/IDA-1.4.4 for ); Tue, 5 Jul 1994 23:36:33 +0200 Message-Id: <199407052136.AA25807@accucx.cc.ruu.nl> Received: by ruucmc.far.ruu.nl (16.6/16.2) id AA07178; Tue, 5 Jul 94 23:34:35 +0200 From: "E.E.Moret" Subject: Summary: S-charges To: chemistry@ruucmc.far.ruu.nl Date: Tue, 5 Jul 94 23:34:34 METDST Name: Ed Moret Organisation: Utrecht University Phone: 030 - 536979 Mailer: Elm [revision: 66.25] Dear colleagues, Thanks to all who responded so helpfully to my question on sensible S-charges. Some replies indicated even higher S-charges, than the ones I calculated semi-empirically. These results were obtained on quite different compounds, however. Some colleagues computed the charges on my model compound with 6-31++G** and also DFT methods. I intend to use these charges, that are close to +1 for the S-atom. For the bioactive molecules we are studying, I will change the semi-empirical charges only on the sulfonamide-moiety accordingly and leave the rest of the molecules (peptides) as computed. A colleague in Leiden is currently working on the ab initio stuff. Like with Mopac-ESP, CHELPG gives charges comparable to Mulliken with 6-31G* basis sets, according to the calculations from Dr. Guadagnini. As his reply shows, however, the CHELPG 6-31++G** S-charge is only +0.5 !! Some references on the derivation of S-charge for force-fields are highly appreciated. Thanks again. The original messages and the replies are appended. I did not include a Dutch reply, offering me a recent CSD-search which would enable me to check the geometry statistics. With best regards, Ed Moret ************************** From: "E.E.Moret" Subject: CCL:Semi-empirical S-charge: 'right' or 'wrong' ? Date: Sun, 26 Jun 94 13:41:40 METDST Dear colleagues, A good scientific discussion is appropriate and let me be the first to say that I am aware of the fact that my subject line is questionable. There is no right or wrong charge. Atomic point charges are merely a useful concept in computational chemistry. However, the semi-empirical charges I have obtained for the hypervalent S atom are a problem. They disrupt molecular mechanics calculations that I want to perform. I face the problem of accepting these charges or considering them useless for my purpose (docking). Here is my dilemma: We are currently studying molecules with around 100 atoms of which a sulfonamide group is of relevance. Apart from a geometry that does not correpond to X-ray and ab initio information, we observe large positive charges on the S atoms. Some calculation results on the molecule CH3-SO2-NH-CH3 follow. The C-S-N-C torsion angle in the input file was close to the ab initio and X-ray data of 70 or -70. I used MINDO3,MNDO,AM1,PM3 Hamiltonians in Ampac 2.1, Mopac 5, Mopac 6 and Mopac 93. The partial charge on the S atom varied between 1.6 and 2.6 ! Fortunately, there was no programme dependency, apart from programmes that explicitly mentioned the use of a different Hamiltonian for one element. This would be a typical result for the S-charge (Mopac93): sulfa_opt_MINDO3: 6 S 1.849085 sulfa_opt_MNDO: 6 S 1.632472 sulfa_opt_AM1: 6 S 2.526724 sulfa_opt_PM3: 6 S 2.159568 The input-file with cartesian coordinates is appended at the bottom. Besides interesting geometry and charge, the bond orders for the bonds from S to C or N are typically 0.7, whereas the bond orders for the bonds from S to O are 1.15. I consider the system size prohibitive for ab initio calculations and I hesitate to use an empirical charge calculation. I would value your input on the following questions: - am I wrong in considering the current results conspicuous? - should I accept the current charges, geometry and bond orders? - should I turn to more rigorous or more empirical methods? - should I fiddle the current charges (i.e. changing the S valency by hand and substracting some pos. charge of S and some neg. charge of its neighbouring atoms, which would be a manual reduction? - are their experimental data which would allow me to check whether the current results predict the right interaction or behaviour of S-containing compounds. - will the P atom give comparable results? For those who noticed that I avoided naming P and S. Is it a phosphorus atom and a sulphurous atom or is sulphur and phospor correct english? I am looking forward to your ideas on this topic. AM1 T=3600 NOINTER PRECISE BONDS ./sulfa_opt_AM1.dat 0001 0.0000 0000 0.0000 0000 0.0000 0000 0000 0000 0000 0001 1.7964 0000 0.0000 0000 0.0000 0000 0000 0000 0000 0001 1.4572 0000 3.3222 0000 0.0000 0000 0000 0000 0000 0001 3.5945 0000 3.7754 0000 0.5259 0000 0000 0000 0000 0006 0.8982 0000 0.1420 0000 0.6190 0000 0000 0000 0000 0016 0.8981 0000 1.7989 0000 1.3224 0000 0000 0000 0000 0007 1.6790 0000 3.0433 0000 1.0195 0000 0000 0000 0000 0006 3.1526 0000 2.9846 0000 1.1500 0000 0000 0000 0000 0001 0.8982 0000 -0.5819 0000 1.4472 0000 0000 0000 0000 0001 3.5287 0000 2.0089 0000 0.8086 0000 0000 0000 0000 0001 3.4511 0000 3.1472 0000 2.1961 0000 0000 0000 0000 0008 0.4152 0000 1.4318 0000 2.6394 0000 0000 0000 0000 0008 -0.3714 0000 2.2528 0000 0.7887 0000 0000 0000 0000 From: "E.E.Moret" Subject: CCL:Semi-empirical S-charges II Date: Mon, 27 Jun 94 13:35:16 METDST Dear colleagues, With regards to the question about 'sensible' charges on sulfur atoms: The first responses to my question have arrived (thanks) and suggest to me that I need to clarify 2 things: - the CH3-SO2-NH-CH3 is only a test model for the sulfonamide containing bioactive compounds with approximately 100 atoms. Large basis set ab initio calculations could only be applied to part of the molecules. - Potential derived charges may be the best for modelling purposes and they may be less basis set dependent. Is this correct, even if the basis set lacks incorporation of d-orbitals? I calculated the ESP (Bessler, Merz, Kollman) charges with Mopac93: Molecule Ch3-SO2-nh-ch3 (important atoms in capital letters): Atom Mulliken ESP C -0.82 -0.99 S 2.53 2.45 N -0.78 -0.96 O -0.84 -0.78 O -0.91 -0.88 ************************** Date: Sun, 26 Jun 94 13:36:22 -0400 From: boehm@mulliken.cem.msu.edu (Randy Boehm) Subject: charges_reply The best (arguably) way to get partial atomic charges is to do a fit to the electrostatic potential. I have done such calculations and have found that S in SO4 is about +3. I your MM applications can not handle such values, then get a differenct or develop your own. ************************** Date: Sun, 26 Jun 94 13:02:45 -0700 From: marvin@biosym.com (Marvin Waldman) Subject: Sulfur charges It is indeed an interesting question as to what is the appropriate charge for a hypervalent sulfur attached to so many electronegative atoms (2 oxygens and a nitrogen). My two suggestions are to see what kinds of charges you can get by fitting to electrostatic potentials, as opposed to the Mulliken charges which are the ones I assume you are reporting. Options for calculating ESP-fitted charges are available in the more recent versions of Ampac and Mopac. Alternatively, can you perform some ab initio calculations (need to be at least at 6-31G* - maybe higher) on a small model compound such as CH3-SO2-NH2 to see what kind of charge is found on the sulfur? I would be interested in seeing the results of any further analysis or prior work that you uncover on this topic posted back to the net. ************************** From: Dan Severance Subject: Re: CCL:Semi-empirical S-charge: 'right' or 'wrong' ? Date: Sun, 26 Jun 94 16:06:46 EDT I don't know if it's an option in MOPAC 93, but try electrostatic potential fit charges; they are MUCH less sensitive to basis set used than Mulliken charges. From: Dan Severance Subject: Re: CCL:Semi-empirical S-charges II Date: Mon, 27 Jun 94 11:38:03 EDT > - Potential derived charges may be the best for modelling purposes and they > may be less basis set dependent. Is this correct, even if the basis set > lacks incorporation of d-orbitals? > I calculated the ESP (Bessler, Merz, Kollman) charges with Mopac93: > Compute the ESP with all 5 basis sets and compare; they are going to be different than the Mulliken charges (in general). We've had surprisingly good luck using CHELPG charges out of G92 (POP=CHELPG,DIPOLE) which constrains the dipole moment to be correct. G92 does have all of the semi empirical basis sets built in to it as well. The other option is to do ab initio (POP=CHELPG or your favorite ESP fit) on this piece just to check the charges for this test molecule. You can always derive the charges on a "per residue" basis approximating that the other pieces of the molecule have a minimal effect on the charges. ************************** Date: Mon, 27 Jun 94 09:34:39 EST From: Dr. Dave Winkler Subject: Re: CCL:Semi-empirical S-charge: 'right' or 'wrong' ? I too have done semiempirical MO calculations on molecules containing sulfur (sulfones and sulfoxides) and have obtained the same behaviour. I would be very interested in what you discover. I had some doubt about the size of the charges but felt that a highly polarizable atom like S, covalently bonded to two oxygens, may have quite a high charge. Please send me any relevant information coming out of this discussion. What about a fixed geometry calculation on something small like DMSO using ab initio methods? ************************** From: "Andrey Bliznyuk" Date: Mon, 27 Jun 1994 17:36:15 Subject: Charges on S It is not clear from your letter, which charges are you going to use. It seems, that you are speaking about Mulliken charges. This charges are not supposed to be used in molecular mechanics calculations, because they do not reproduce electrostatic potential around your molecule. The appropriate way is to use charges generated from electrostatic potential calculation. Most programs can do that. Recently, interesting method have been introduced: G.P. Ford, B. Wang - J. Comp. Chem., 1994, v.15, p.200-207. Ordinary, using such methods you receive hypervalent charges for S / P around 1.8. The problem is that hypervalent S and P are unaccessible. There is a way around: C.I. Bayly, P. Cieplak, W.D. Cornell, P.A. Kollman: J. Phys. Chem., 1993, v.97. p.10269-10280. but, as far as i know, it is not implemented in Ampack/Mopack. Fortunately, for most of the calculations, it is not very important, because the electrostatic potential around molecule is good enough. My experience (for P) is that AM1 charges, derived from electrostatic calculations are good. The Mulliken charges looks ok for me, for example for H2SO4: MNDO = 1.7483 - not parametrizied for hypervalent AM1 = 2.8734 PM3 = 2.5095 MNDO/d = 2.3401 - d orbitals on S. P.S. There are some confusing about Mulliken charges in semiempirical calculations. Here, i am using the standard (in NDO approximation) definition: q(atom a) = Sum(i of a) Pii, where Pii is diagonal elements of density matrix. ************************** Date: Mon, 27 Jun 1994 10:27:50 +0200 From: kit1@mri.ch.cam.ac.uk Subject: sulphur things I just read your email to CCL on sulphur cpds. I tried CHARGE3 in XED (as I expect you have anyway) and found that the charge on the S was +0.2937 . I realise that CHARGE3 is empirical, but MOPAC is semi-empirical. with differences between Hamiltonians. ...stuff deleted... Even when using ab initio ESPs the situation is not clear cut as there is also a basis set dependence. I think you should use the charges you think give the 'best' results, as long as you are consistent in the use of charge models and use a 'standard' of some sort (not much of a problem as there are lots of standards to choose from). I believe the correct international (IUPAC) name for sulphur is sulfur, but I don't think phosphorus was 'changed' for the benefit of 'global communications'. ************************** Date: Mon, 27 Jun 94 10:50:26 +0100 From: "Wojciech Galazka" Subject: Re: CCL:Semi-empirical S-charge: 'right' or 'wrong' ? And did you include d-orbitals to your calculations? Look for SIBIQ program that uses MNDO/d scheme. ************************** Date: Mon, 27 Jun 1994 12:03:24 -0400 (EDT) From: GKING@arserrc.gov Subject: Re: CCL:Semi-empirical S-charge: 'right' or 'wrong' ? I am also interested in the "correct" atomic charges one should use for the sulfone group. I do not have access to programs like G92, so I am interested in your results. Have you performed any G92 calculations on your model molecule CH3-SO2-NH-CH3, or just semi-empirical calculations? A few years ago I found a paper in which the charge of the S in the sulfone group was apparently found to be about 0.9 through some sort of spectroscopy that I am not familiar with. I'm not positive about this number, because the paper was written in Russian (which I do not understand). The number 0.9 appeared in a table in this paper, so I may or may not be interpreting it correctly. I can probably find this paper in my files and provide you with the reference if you are interested. ************************** From: "Christian Boehme" Date: Mon, 27 Jun 1994 18:41:01 MDT Subject: Re: CCL:Semi-empirical S-charge: 'right' or 'wrong' ? To me your results sound reasonable. If you use a semipolar bonding model for this hypervalent compound (which is preferred by a lot of theoretical chemists instead of d-orbital participation), the charge should be around 2.0. I obtained the following Sulfur charges for SF6, using an 6-31G* basis on RHF level : Mulliken : 2.4 Natural Population Analysis : 3.2 Bader s Atoms In Molecules : 4.5 Semipolar bonding suggests a charge around 3.0. So the higher values you obtained may even be the more reasonable ones. Of course you re right in your opinion that atom charges are only a model. I ve recently read a reference which compared different methods to derive such charges. I don t have it at hand but if you re interested, just send me a note. Greetings, ************************** Date: 27 Jun 94 14:21:00 EST From: "DR. LIOTTA" Subject: sulfonamide (repeat message - new address found) I only caught the second message you sent to the CCL but I am very interested in your test model for the sulfonamide bioactive compound. I am trying to model a similar compound, but need to find the v1, v2 and v3 torsional vibration parameters for -S(O2)-N(CH3)-C(sp3)-OH, -S(O2)-N(CH3)-C(sp3)-C(sp2) and -S(O2)-N(CH3)-C(sp3)-C(sp3) Have these been calculated before and if so could you tell me where they are published. Thank you. Veronica Cornel ************************** Date: Tue, 28 Jun 1994 09:52:13 -0400 From: enrico.purisima@BRI.NRC.CA (Enrico Purisima) Subject: CCL:Sulfonamide Two papers that derive force-field parameters for sulfonamides are: Nicholas et al., "A Molecular Mechanics Valence Force Field for Sulfonamides Derived by Ab Initio Methods." J. Phys. Chem. 95:9803-9811 (1991). Bindal et al., "Ab Initio Calculations on N-Methylmethansulfonamide and Methyl Methanesulfonate for the Development of Force Field Parameters and Their Use in the Conformational Analysis of Some Novel Estrogens." JACS 112:7861-7868 (1990). ************************** From: Paulo Henrique Guadanini Subject: S-charges Date: Mon, 27 Jun 1994 17:18:30 -0300 (EST) Hi Moret, I calculated charges for your model compound (CH3-SO2-NH-CH3) using 2 ab initio wave functions (table) . Note that S ESP charge for HF/6-31++G** wave function is 2.5 times smaller than others theoretically derived charges. I included also charges calculated more empirically by the Charge Equilibration method (A.K.Rappe and W.A.Goddard III, J.Phys.Chem. 95,3358(1991)). In this case the S charge have opposite sign. Due to the modest computational costs, you may try to use methods based on the electronegativity equalization approach to calculate charges for the bioactive compounds. See, for example J.Mullay J.Comp.Chem. 9(7), 764(1988) J.Mullay J.Comp.Chem. 12(3), 369(1991) W.J.Mortier, et al J.Am.Chem.Soc. 107,829(1985) W.J.Mortier, et al J.Am.Chem.Soc. 108,4315(1986) T.K.Ghanty and S.K.Ghosh THEOCHEM 276, 83(1992) J.Gasteiger and M.Marsili Tetrahedron 36,3219(1980) A.K.Rappe and W.A.Goddard III, J.Phys.Chem. 95,3358(1991) HF/6-31G HF/6-31G HF/6-31++G** HF/6-31++G** Charge Mulliken CHELPG Mulliken CHELPG Equilibration H .175192 -.063406 .171631 -.083289 .15623 H .052421 -.168608 .006913 -.208642 .08554 H .540226 .371062 .501927 .392285 .26842 H .211525 .067218 .168815 .062701 .11286 C -.949472 -.111049 -.752617 .114410 -.07299 S 1.230307 1.166504 1.433971 .519700** -.09075** N -.514572 -.065378 -.446039 .086295 -.24547 C -.182035 -.092049 -.252464 -.026867 -.20924 H .232512 .007172 .174369 -.029782 .15315 H .243595 .158310 .187150 .151572 .13459 H .255879 .104173 .210553 .095389 .17736 O -.677410 -.694042 -.767149 -.552283 -.21774 O -.618167 -.679908 -.637061 -.521488 -.25194 ************************** From: "Chris L. Waller" Date: Fri, 1 Jul 1994 10:17:12 -0400 Subject: Re: CCL:Semi-empirical S-charges II About sulfonamides: I ran a DFT calculation on your test molecule using a DNP (double numerical with polarization functions) basis set and obtained the following charges: Effective atomic charges from Mulliken populations: O # 1 -0.457 S # 2 0.929 O # 3 -0.455 C # 4 -1.092 N # 5 -0.652 C # 6 -0.906 H # 7 0.362 H # 8 0.388 H # 9 0.395 H # 10 0.460 H # 11 0.354 H # 12 0.337 H # 13 0.336 This was a single point energy calculation on the PM3 optimized structure. ************************** End of the replies. ------------------------------------------------------------------------- E.E. Moret (@more@) E.E.Moret@far.ruu.nl Computational Medicinal Chemistry/Department of Pharmaceutical Chemistry Faculty of Pharmacy/Utrecht University/the Netherlands Telephone (31-30)536979/536958 Facsimile (31-30)516674 -------------------------------------------------------------------------