Skip to content

Research at St Andrews

Can human experts predict solubility better than computers?

Research output: Contribution to journalArticle

Author(s)

Samuel Boobier, Anne Osbourn, John B. O. Mitchell

School/Research organisations

Abstract

In this study, we design and carry out a survey, asking human experts to predict the aqueous solubility of druglike organic compounds. We investigate whether these experts, drawn largely from the pharmaceutical industry and academia, can match or exceed the predictive power of algorithms. Alongside this, we implement 10 typical machine learning algorithms on the same dataset. The best algorithm, a variety of neural network known as a multi-layer perceptron, gave an RMSE of 0.985 log S units and an R2 of 0.706. We would not have predicted the relative success of this particular algorithm in advance. We found that the best individual human predictor generated an almost identical prediction quality with an RMSE of 0.942 log S units and an R2 of 0.723. The collection of algorithms contained a higher proportion of reasonably good predictors, nine out of ten compared with around half of the humans. We found that, for either humans or algorithms, combining individual predictions into a consensus predictor by taking their median generated excellent predictivity. While our consensus human predictor achieved very slightly better headline figures on various statistical measures, the difference between it and the consensus machine learning predictor was both small and statistically insignificant. We conclude that human experts can predict the aqueous solubility of druglike molecules essentially equally well as machine learning algorithms. We find that, for either humans or algorithms, combining individual predictions into a consensus predictor by taking their median is a powerful way of benefitting from the wisdom of crowds.
Close

Details

Original languageEnglish
JournalJournal of Cheminformatics
Volume9
Issue number63
DOIs
StatePublished - 13 Dec 2017

Discover related content
Find related publications, people, projects and more using interactive charts.

View graph of relations

Related by author

  1. Crystal structure evaluation: calculating relative stabilities and other criteria: general discussion

    Addicoat, M. , Adjiman, C. S. , Arhangelskis, M. , Beran, G. J. O. , Bowskill, D. , Brandenburg, J. G. , Braun, D. E. , Burger, V. , Cole, J. , Cruz-Cabeza, A. J. , Day, G. M. , Deringer, V. L. , Guo, R. , Hare, A. , Helfferich, J. , Hoja, J. , Iuzzolino, L. , Jobbins, S. , Marom, N. , McKay, D. & 21 others Mitchell, J. B. O., Mohamed, S., Neumann, M., Lill, S. N., Nyman, J., Oganov, A. R., Piaggi, P., Price, S. L., Reutzel-Edens, S., Rietveld, I., Ruggiero, M., Ryder, M. R., Sastre, G., Schön, J. C., Taylor, C., Tkatchenko, A., Tsuzuki, S., van den Ende, J., Woodley, S. M., Woollam, G. & Zhu, Q. 10 Oct 2018 In : Faraday Discussions. 211, p. 325-381

    Research output: Contribution to journalComment/debate

  2. Applications of crystal structure prediction – inorganic and network structures: general discussion

    Burger, V., Claeyssens, F., Davies, D. W., Day, G. M., Dyer, M. S., Hare, A., Li, Y., Mellot-draznieks, C., Mitchell, J. B. O., Mohamed, S., Oganov, A. R., Price, S. L., Ruggiero, M., Ryder, M. R., Sastre, G., Schön, J. C., Spackman, P., Woodley, S. M. & Zhu, Q. 9 Oct 2018 In : Faraday Discussions. 211, p. 613-642

    Research output: Contribution to journalComment/debate

  3. Applications of crystal structure prediction – organic molecular structures: general discussion

    Adjiman, C. , Brandenburg, J. G. , Braun, D. E. , Cole, J. , Collins, C. , Cooper, A. I. , Cruz-Cabeza, A. , Day, G. , Dudek, M. , Hare, A. , Iuzzolino, L. , McKay, D. , Mitchell, J. B. O. , Mohamed, S. , Neelamraju, S. , Neumann, M. , Lill, S. N. , Nyman, J. , Oganov, A. R. , Price, S. L. & 9 others Pulido, A., Reutzel-Edens, S., Rietveld, I., Ruggiero, M. T., Schön, J. C., Tsuzuki, S., van den Ende, J., Woollam, G. & Zhu, Q. 4 Oct 2018 In : Faraday Discussions. 211, p. 493-539

    Research output: Contribution to journalComment/debate

  4. Artificial intelligence in pharmaceutical research and development

    Mitchell, J. B. O. Jul 2018 In : Future Medicinal Chemistry. 10, 13, p. 1529-1531 3 p.

    Research output: Contribution to journalEditorial

  5. Enzyme function and its evolution

    Mitchell, J. B. O. Dec 2017 In : Current Opinion in Structural Biology. 47, p. 151-156

    Research output: Contribution to journalReview article

Related by journal

  1. A note on utilising binary features as ligand descriptors

    Mussa, H. Y., Mitchell, J. B. O. & Glen, R. 1 Dec 2015 In : Journal of Cheminformatics. 7, 58

    Research output: Contribution to journalArticle

  2. Verifying the fully “Laplacianised” posterior Naïve Bayesian approach and more

    Mussa, H. Y., Marcus, D., Mitchell, J. B. O. & Glen, R. 12 Jun 2015 In : Journal of Cheminformatics. 7, 27

    Research output: Contribution to journalArticle

  3. Full "laplacianised" posterior naive Bayesian algorithm

    Mussa, H. Y., Mitchell, J. B. O. & Glen, R. C. 23 Aug 2013 In : Journal of Cheminformatics. 5, 8, 37

    Research output: Contribution to journalArticle

  4. Predicting the protein targets for athletic performance-enhancing substances

    Mavridis, L. & Mitchell, J. B. O. 25 Jun 2013 In : Journal of Cheminformatics. 5, 31, 13 p., 31

    Research output: Contribution to journalArticle

ID: 251536334