1. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Properties predicted by COSMO-RS-based QSAR Dr. Karin Wichmann COSMOlogic GmbH & Co. KG Burscheider Str. 515 D-51381 Leverkusen Germany Phone: +49-2171-731-683 Fax: +49-2171-731-689 Email: wichmann@cosmologic.de Web: http://www.cosmologic.de

2. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Property Prediction with the -potential COSMOtherm allows for the prediction of the entire fluid phase equilibrium thermodynamics of almost arbitrary compounds, e.g: -Activity coefficients -Solubility -Partition coefficients between well characterized liquid phases The chemical potential and thermodynamic properties of liquids can easily be computed using the COSMO-RS -potential S ( ): e.g. activity coefficients X S :

3. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Treatment of complex or chemically less well-defined matrices Some physiological and thermodynamic equilibria involve phases or matrices which are -Chemically not well defined or of unknown composition -Disordered -Not really liquid / amorphous E.g. phases like -Biological tissues, blood, brain, the intestinal resorption system -Soap, polymers, cotton -Solid adsorbents like activated carbon -potential cannot be calculated directly by COSMO-RS

4. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Treatment of complex or chemically less well-defined matrices Chemically not well defined phases can be treated indirectly with the -moments: The -potential can be approximated very accurately by a set of functions M i X : -moments M i X are defined with: And Hydrogen Bonding moments M i,HB X M 0 X : molecular area M 1 X : negative total charge of compound X M 2 X : correlated with the screening energy for X M 3 X : skewness of the profile M -1/-2 X or M don/acc X : hydrogen bond capacity

5. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Treatment of complex or chemically less well-defined matrices With the -moments, -each phase / matrix can be characterized by a set of -coefficients c i S. -Differences between two phases can be described by the same kind of expansion, with coefficients c i S,S being the difference of the coefficients of the two phases. -Partition coefficients between two phases are generally expressed by a linear combination of the compounds -moments in the form:

6. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 QSPR applications -moments can be used as descriptors for linear QSAR/QSPR applications. Using COSMOtherm -moment descriptors it is possible to predict: -Partition behaviour between almost arbitrary phases, such as -CaCo2 cell permeability -logK BB blood-brain barrier partition -logK OC soil-water partitioning (environmental fate) -logK IA intestinal absorption -... and many more -Solubility of drugs in arbitrary solvents -Chemical potentials of crystal surfaces in solution morphology of drugs

7. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 QSPR applications -Only five -moments M i X are necessary to get a general QSPR description of solvents or biological matrices: -This is in correspondence with the general solvation theory of Abraham, which claims that solvent space is five-dimensional. -Intercorrelation of the COSMOtherm -moments and the Abraham set of solvent descriptors is possible* *Andreas M. Zissimos, Michael H. Abraham, Andreas Klamt, Frank Eckert, John Wood, J. Chem. Inf. Comput. Sci. 42 (2002) 1320

8. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Blood-Brain Partitioning logBB = 0.0043 M 0 - 0.017 M 2 - 0.0031 M 3 + 0.24 N = 103 rms = 0.41 R 2 = 0.691 Exp. Data: Rose, K.; Hall, L. H.; Hall, L. M.; Kier, L. B. Modeling Blood-Brain Barrier Partitioning Using Topological Structure Descriptors, 2003. Elsevier MDL Web Site. http://www.mdl.com/solutions/white_papers/MDLQSARreprint.jsp (accessed Feb 13, 2004).

9. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Human Serum Albumin Binding logK HSA = 0.0084 M 0 - 0.017 M 2 - 0.0014 M 3 + 0.17 M 3,acc + 0.24 N = 94 rms = 0.33 R 2 = 0.671 Exp. Data: Kier, L. B.; Hall, L. H.; Hall, L. M. QSAR Modeling of Drug Binding to Protein. -Lactam Serum Binding and Albumin Binding Affinity, 2002. Elsevier MDL Web Site. http://www.mdl.com/solutions/white_papers/ML233QM3_white_paper.jsp (accessed Oct 7, 2004).

10. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Intestinal Absorption logK HSA = 0.0040 M 0 - 0.0053 M 2 - 0.0024 M 3 - 0.11 M 3,acc - 0.11 M 3,don + 1.37

11. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Soil Sorption logK OC = 0.016 M 0 - 0.017 M 2 - 0.046 M 3 + 0.23 M 3,acc – 0.33 M 3,don + 0.41 N = 386 rms = 0.61 R 2 = 0.72

12. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Adsorption to Activated Carbon [Mehler, Peukert (TU München), Klamt; AICHE Journal. 48, 1093-1099 (2002)]

13. ©2009, COSMOlogic GmbH & Co KGCOSMO-RS Symposium 2009 Model of Cytochrome P450 2D6 and Cytochrome P450 3A4 Modelling inhibition of Cytochrome P450 2D6 using -moment descriptors for a diverse set of 33 compounds from the NCI database. Log(IC50) = 2.96(0.56) – 0.0097(0.0018)*Area + 0.010(0.0067)*sig2 + 0.050(0.012)*sig3 - 0.42(0.089)*HBacc3 + 0.39(0.14)*HBdon3 Modelling inhibition of Cytochrome P450 3A4 using -moment descriptors for 45 structurally diverse drugs. (2 outliers deleted.) Log(IC50) = 4.46(0.50) – 0.0093(0.0017)*Area - 0.0068(0.0078)*sig2 + 0.025(0.010)*sig3 + 0.033(0.128)*HBacc3 + 0.24(0.14)*HBdon3 Dataset from Pharm. Res. 2003, 20, 1401.