Journal of Computer-Aided Molecular Design

, Volume 32, Issue 10, pp 1165–1177 | Cite as

SAMPL6 challenge results from \(pK_a\) predictions based on a general Gaussian process model

  • Caitlin C. Bannan
  • David L. Mobley
  • A. Geoffrey SkillmanEmail author


A variety of fields would benefit from accurate \(pK_a\) predictions, especially drug design due to the effect a change in ionization state can have on a molecule’s physiochemical properties. Participants in the recent SAMPL6 blind challenge were asked to submit predictions for microscopic and macroscopic \(pK_a\)s of 24 drug like small molecules. We recently built a general model for predicting \(pK_a\)s using a Gaussian process regression trained using physical and chemical features of each ionizable group. Our pipeline takes a molecular graph and uses the OpenEye Toolkits to calculate features describing the removal of a proton. These features are fed into a Scikit-learn Gaussian process to predict microscopic \(pK_a\)s which are then used to analytically determine macroscopic \(pK_a\)s. Our Gaussian process is trained on a set of 2700 macroscopic \(pK_a\)s from monoprotic and select diprotic molecules. Here, we share our results for microscopic and macroscopic predictions in the SAMPL6 challenge. Overall, we ranked in the middle of the pack compared to other participants, but our fairly good agreement with experiment is still promising considering the challenge molecules are chemically diverse and often polyprotic while our training set is predominately monoprotic. Of particular importance to us when building this model was to include an uncertainty estimate based on the chemistry of the molecule that would reflect the likely accuracy of our prediction. Our model reports large uncertainties for the molecules that appear to have chemistry outside our domain of applicability, along with good agreement in quantile–quantile plots, indicating it can predict its own accuracy. The challenge highlighted a variety of means to improve our model, including adding more polyprotic molecules to our training set and more carefully considering what functional groups we do or do not identify as ionizable.


\(pK_a\) SAMPL6 Blind challenge Gaussian process 



DLM and CCB appreciate the financial support from the National Science Foundation (CHE 1352608) and the National Institutes of Health (1R01GM108889-01). CCB was supported financially by OpenEye Scientific Software to build this model during Summer 2017 and is now supported by a fellowship from The Molecular Sciences Software Institute under NSF Grant ACI-1547580. We are thankful for valuable conversations with OpenEye employees, the SAMPL6 organizers, and all challenge participants, and especially to Merck for its contributions to the experimental work in this challenge. AGS would like to thank Paul Hawkins, Christopher Bayly and Robert Tolbert as well as Anthony Nicholls and Matthew Geballe for many insightful discussions of \(pK_a\) and machine learning.

Supplementary material

10822_2018_169_MOESM1_ESM.pdf (706 kb)
Supplementary material 1 (PDF 707 KB)
10822_2018_169_MOESM2_ESM.tar (7.3 mb)
Supplementary material 1 (TAR 7429 KB)


  1. 1.
    Wan H, Ulander J (2006) High-throughput pKa screening and prediction amenable for ADME profiling. Expert Opin Drug Metab Toxicol 2(1):139. CrossRefPubMedGoogle Scholar
  2. 2.
    Gleeson MP (2008) Generation of a set of simple, interpretable ADMET rules of thumb. J Med Chem 51(4):817. CrossRefPubMedGoogle Scholar
  3. 3.
    Manallack DT, Prankerd RJ, Yuriev E, Oprea TI, Chalmers DK (2013) The significance of acid/base properties in drug discovery. Chem Soc Rev 42(2):485. CrossRefPubMedPubMedCentralGoogle Scholar
  4. 4.
    Manchester J, Walkup G, Rivin O, You Z (2010) Evaluation of pKa estimation methods on 211 druglike compounds. J Chem Inf Model 50(4):565. CrossRefPubMedGoogle Scholar
  5. 5.
    Settimo L, Bellman K, Knegtel RMA (2014) Comparison of the accuracy of experimental and predicted pKa values of basic and acidic compounds. Pharm Res 31(4):1082. CrossRefPubMedGoogle Scholar
  6. 6.
    Fraczkiewicz R (2013) In silico prediction of ionization. In: Reedijk J (ed) Reference module in chemistry, molecular sciences and chemical engineering. Elsevier, WalthamGoogle Scholar
  7. 7.
    Bannan CC, Burley KH, Chiu M, Shirts MR, Gilson MK, Mobley DL (2016) Blind prediction of cyclohexane–water distribution coefficients from the SAMPL5 challenge. J Comput Aided Mol Des 30(11):1. CrossRefGoogle Scholar
  8. 8.
    Pickard FC, König G, Tofoleanu F, Lee J, Simmonett AC, Shao Y, Ponder JW, Brooks BR (2016) Blind prediction of distribution in the SAMPL5 challenge with QM based protomer and pKa corrections. J Comput Aided Mol Des 30(11):1. CrossRefGoogle Scholar
  9. 9.
    Aguilar B, Anandakrishnan R, Ruscio JZ, Onufriev AV (2010) Statistics and physical origins of pK and ionization state changes upon protein-ligand binding. Biophys J 98(5):872. CrossRefPubMedPubMedCentralGoogle Scholar
  10. 10.
    Işık M, Levorse D, Rustenburg AS, Ndukwe IE, Wang H, Wang X, Reibarkh M, Martin GE, Makarov AA, Mobley DL, Rhodes T, Chodera JD (2018) pka measurements for the sampl6 prediction challenge for a set of kinase inhibitor-like fragments. bioRxiv.
  11. 11.
    Darvey IG (1995) The assignment of pKa values to functional groups in amino acids. Biochem Educ 23(2):80. CrossRefGoogle Scholar
  12. 12.
    Bodner GM (1986) Assigning the pKa’s of polyprotic acids. J Chem Educ 63(3):246. CrossRefGoogle Scholar
  13. 13.
    Işık M, Rustenburg AS (2018) Michael, Shirts, D.L. Mobley, J.D. Chodera. SAMPL6.
  14. 14.
    Exner O (1972) Advances in linear free energy relationships. Springer, Boston. CrossRefGoogle Scholar
  15. 15.
    Perrin D, Dempsey B, Serjeant E (1981) pKa prediction for organic acids and bases. Chapman and Hall, New YorkCrossRefGoogle Scholar
  16. 16.
    Geidl S, Svobodová Vařeková R, Bendová V, Petrusek L, Ionescu CM, Jurka Z, Abagyan R, Koča J (2015) How does the methodology of 3D structure preparation influence the quality of pKa prediction? J Chem Inf Model 55(6):1088. CrossRefPubMedPubMedCentralGoogle Scholar
  17. 17.
    Cruciani G, Milletti F, Storchi L, Sforna G, Goracci L (2009) In silico pKa prediction and ADME profiling. Chem Biodivers. 6(11):1812. CrossRefPubMedGoogle Scholar
  18. 18.
    Katritzky AR, Kuanar M, Slavov S, Hall CD, Karelson M, Kahn I, Dobchev DA (2010) Quantitative correlation of physical and chemical properties with chemical structure: utility for prediction. Chem Rev 110(10):5714. CrossRefPubMedGoogle Scholar
  19. 19.
    Peterson KL (2000) Reviews in computational chemistry. Wiley, HobokenGoogle Scholar
  20. 20.
    Fraczkiewicz R, Lobell M, Göller AH, Krenz U, Schoenneis R, Clark RD, Hillisch A (2015) Best of both worlds: combining pharma data and state of the art modeling technology to improve in silico pKa prediction. J Chem Inf Model 55(2):389. CrossRefPubMedGoogle Scholar
  21. 21.
    Citra MJ (1999) Estimating the pKa of phenols, carboxylic acids and alcohols from semi-empirical quantum chemical methods. Chemosphere 38(1):191. CrossRefPubMedGoogle Scholar
  22. 22.
    Vařeková RS, Geidl S, Ionescu CM, Skřehota O, Bouchal T, Sehnal D, Abagyan R, Koča J (2013) Predicting pKa values from EEM atomic charges. J Cheminf 5:18. CrossRefGoogle Scholar
  23. 23.
    Dixon SL, Jurs PC (1993) Estimation of pKa for organic oxyacids using calculated atomic charges. J Comput Chem 14(12):1460. CrossRefGoogle Scholar
  24. 24.
    Zevatskii YE, Samoilov DV (2011) Modern methods for estimation of ionization constants of organic compounds in solution. Russ J Org Chem 47(10):1445. CrossRefGoogle Scholar
  25. 25.
    Pracht P, Bauer CA, Grimme S (2017) Automated and efficient quantum chemical determination and energetic ranking of molecular protonation sites. J Comput Chem 38(30):2618. CrossRefPubMedGoogle Scholar
  26. 26.
    Bochevarov AD, Harder E, Hughes TF, Greenwood JR, Braden DA, Philipp DM, Rinaldo D, Halls MD, Zhang J, Friesner RA, Jaguar (2013) A high-performance quantum chemistry software program with strengths in life and materials sciences. Int J Quantum Chem 113(18):2110. CrossRefGoogle Scholar
  27. 27.
    Bochevarov AD, Watson MA, Greenwood JR, Philipp DM (2016) Multiconformation, density functional theory-based pka prediction in application to large, flexible organic molecules with diverse functional groups. J Chem Theory Comput 12(12):6001. CrossRefGoogle Scholar
  28. 28.
    Rasmussen CE, Williams CKI (2006) Gaussian processes for machine learning, adaptive computation and machine learning. MIT Press, CambridgeGoogle Scholar
  29. 29.
    OpeneEye Scientific Software, Inc. OEChem Toolkit (2018).
  30. 30.
    Hawkins PCD, Skillman AG, Warren GL, Ellingson BA, Stahl MT (2010) Conformer generation with OMEGA: algorithm and validation using high quality structures from the protein databank and cambridge structural database. J Chem Inf Model 50(4):572. CrossRefPubMedPubMedCentralGoogle Scholar
  31. 31.
    Wiberg KB (1968) Application of the pople-santry-segal CNDO method to the cyclopropylcarbinyl and cyclobutyl cation and to bicyclobutane. Tetrahedron 24(3):1083. CrossRefGoogle Scholar
  32. 32.
    Mayer I (2007) Bond order and valence indices: a personal account. J Comput Chem 28(1):204. CrossRefPubMedGoogle Scholar
  33. 33.
    OpeneEye Scientific Software, Inc. OEQuacPac Toolkit (2018).
  34. 34.
    Jakalian A, Bush BL, Jack DB, Bayly CI (2000) Fast, efficient generation of high-quality atomic charges. AM1-BCC model: I. Method. J Comput Chem 21(2):132CrossRefGoogle Scholar
  35. 35.
    Jakalian A, Jack DB, Bayly CI (2002) Fast, efficient generation of high-quality atomic charges. AM1-BCC model: II. Parameterization and validation. J Comput Chem 23(16):1623. CrossRefGoogle Scholar
  36. 36.
    Jelfs S, Ertl P, Selzer P (2007) Estimation of pKa for druglike compounds using semiempirical and information-based descriptors. J Chem Inf Model 47(2):450. CrossRefPubMedGoogle Scholar
  37. 37.
    Nicholls A, Wlodek S, Grant JA (2010) SAMPL2 and continuum modeling. J Comput Aided Mol Des 24(4):293. CrossRefPubMedGoogle Scholar
  38. 38.
    Grant JA, Pickup BT, Nicholls A (2001) A smooth permittivity function for Poisson-Boltzmann solvation methods. J Comput Chem 22(6):608. CrossRefGoogle Scholar
  39. 39.
    Nicholls A (2004) Spicoli: a surface toolkit, dudeGoogle Scholar
  40. 40.
    Lee B, Richards FM (1971) The interpretation of protein structures: estimation of static accessibility. J Mol Biol 55(3):379. CrossRefPubMedGoogle Scholar
  41. 41.
    Connolly ML (1983) Analytical molecular surface calculation. J Appl Cryst 16(5):548. CrossRefGoogle Scholar
  42. 42.
    Sharp KA, Nicholls A, Fine RF, Honig B (1991) Reconciling the magnitude of the microscopic and macroscopic hydrophobic effects. Science 252(5002):106. CrossRefPubMedGoogle Scholar
  43. 43.
    Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V, Vanderplas J, Passos A, Cournapeau D, Brucher M, Perrot M, Duchesnay E (2011) Scikit-learn: machine learning in python. J Mach Learn Res 12:2825Google Scholar
  44. 44.
    Kortüm G, Vogel W, Andrussow K (1960) Disssociation constants of organic acids in aqueous solution. Pure Appl Chem 1(2–3):187. CrossRefGoogle Scholar
  45. 45.
    Perrin DD (1972) Dissociation constants of organic bases in aqueous solution: supplement 1972. Butterworths, LondonGoogle Scholar
  46. 46.
    Serjeant P, Dempsey B (1979) Ionisation constants of organic acids in aqueous solution. Pergamon, OxfordGoogle Scholar
  47. 47.
    Hastie T, Tibshirani R, Friedman JH (2009) The elements of statistical learning: data mining, inference, and prediction, 2nd edn. Springer, New YorkCrossRefGoogle Scholar
  48. 48.
    Kuhn HW (2004) The Hungarian method for the assignment problem. Nav Res Logist 52(1):7. CrossRefGoogle Scholar
  49. 49.
    Advanced Chemistry Development, Inc. pKa GALAS (2015).
  50. 50.
  51. 51.
    Goldfarb AR, Mele A, Gutstein N (1955) Basicity of the amide bond. J Am Chem Soc 77(23):6194. CrossRefGoogle Scholar
  52. 52.
    Bordwell FG, Algrim DJ, Harrelson JA (1988) The relative ease of removing a proton, a hydrogen atom, or an electron from carboxamides versus thiocarboxamides. J Am Chem Soc 110(17):5903. CrossRefGoogle Scholar
  53. 53.
    Evans RE (1964) 460. hydropyrimidines. part iii. reduction of amino-pyrimidines. J Chem Soc. CrossRefGoogle Scholar
  54. 54.
    Mobley DL, Wymer KL, Lim NM, Guthrie JP (2014) Blind prediction of solvation free energies from the SAMPL4 challenge. J Comput Aided Mol Des 28(3):135. CrossRefPubMedPubMedCentralGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2018

Authors and Affiliations

  1. 1.Department of ChemistryUniversity of CaliforniaIrvineUSA
  2. 2.2017 Summer Intern, OpenEye Scientific Software, Inc.Santa FeUSA
  3. 3.Departments of Pharmaceutical Sciences and ChemistryUniversity of CaliforniaIrvineUSA
  4. 4.OpenEye Scientific Software, Inc.Santa FeUSA

Personalised recommendations