Reliable Prediction Errors for Deep Neural Networks Using Test-Time Dropout
Open accessAbstract
While the use of deep learning in drug discovery is gaining increasing attention, the lack of methods to compute reliable errors in prediction for Neural Networks prevents their application to guide decision making in domains where identifying unreliable predictions is essential, e.g., precision medicine. Here, we present a framework to compute reliable errors in prediction for Neural Networks using Test-Time Dropout and Conformal Prediction. Specifically, the algorithm consists of training a single Neural Network using dropout, and then applying it N times to both the validation and test sets, also employing dropout in this step. Therefore, for each instance in the validation and test sets an ensemble of predictions are generated. The residuals and absolute errors in prediction for the validation set are then used to compute prediction errors for the test set instances using Conformal Prediction. We show using 24 bioactivity data sets from ChEMBL 23 that Dropout Conformal Predictors are valid (i.e., the fraction of instances whose true value lies within the predicted interval strongly correlates with the confidence level) and efficient, as the predicted confidence intervals span a narrower set of values than those computed with Conformal Predictors generated using Random Forest (RF) models. Lastly, we show in retrospective virtual screening experiments that dropout and RF-based Conformal Predictors lead to comparable retrieval rates of active compounds. Overall, we propose a computationally efficient framework (as only N extra forward passes are required in addition to training a single network) to harness Test-Time Dropout and the Conformal Prediction framework, which is generally applicable to generate reliable prediction errors for Deep Neural Networks in drug discovery and beyond.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
We assigned an applicability score (α) to predictions based on the proximity of a predicted signature to true signatures available in the training set, the robustness of the SNN output to a test-time data dropout, and the accuracy expected a priori based on the experimental CC data sets available for the molecule.
“To explore the reliability of the inferred signatures, we assigned an applicability score (α) to predictions based on the following: (a) the proximity of a predicted signature to true (experimental) signatures available in the training set; (b) the robustness of the SNN output to a test-time data dropout12; and (c) the accuracy expected a priori based on the experimental CC data sets available for”
From Bioactivity descriptors for uncharacterized chemical compounds (2021) · cites this paper as [12]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538 · see citing works
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126 · see citing works
Scikit-learn: Machine Learning in Python
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Müller, A., Nothman, J., Louppe, G., et al. (2012) · arXiv · doi:10.48550/arxiv.1201.0490 · Cited by 71,016 · see citing works
(2014) · Cited by 33,419 · see citing works
(2017) · Cited by 11,625 · see citing works
Extended-Connectivity Fingerprints
Rogers, D. O. W., & Hahn, M. (2010) · Journal of Chemical Information and Modeling · doi:10.1021/ci100050t · Cited by 7,597 · see citing works
Dropout as a bayesian approximation: Representing model uncertainty in deep learning
Gal, Y., & Ghahramani, Z. (2015) · arXiv · doi:10.48550/arxiv.1506.02142 · Cited by 6,471 · see citing works
ChEMBL: a large-scale bioactivity database for drug discovery
Gaulton, A., Bellis, L. J., Bento, A. P., Chambers, J., Davies, M., Hersey, A., Light, Y., McGlinchey, S., Michalovich, D., Al-Lazikani, B., et al. (2012) · Nucleic Acids Research · doi:10.1093/nar/gkr777 · Cited by 4,382 · see citing works
(2013) · Cited by 3,740 · see citing works
What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?
Kendall, A., & Gal, Y. (2017) · arXiv · doi:10.48550/arxiv.1703.04977 · Cited by 3,088 · see citing works
Cyclical Learning Rates for Training Neural Networks
Smith, L. N. (2017) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv.2017.58 · Cited by 2,931 · see citing works
Coudray, N., Ocampo, P. S., Sakellaropoulos, T., Narula, N., Snuderl, M., Fenyö, D., Moreira, A. L., Razavian, N., & Tsirigos, A. (2018) · Nature Medicine · doi:10.1038/s41591-018-0177-5 · Cited by 2,930 · see citing works
Artificial intelligence in healthcare
Yu, K.-H., Beam, A. L., & Kohane, I. S. (2018) · Nature Biomedical Engineering · doi:10.1038/s41551-018-0305-z · Cited by 2,817 · see citing works
Gaulton, A., Hersey, A., Nowotka, M., Bento, A. P., Chambers, J. K., Méndez, D., Mutowo, P., Atkinson, F. L., Bellis, L. J., Cibrián–Uhalte, E., et al. (2016) · Nucleic Acids Research · doi:10.1093/nar/gkw1074 · Cited by 2,351 · see citing works
The rise of deep learning in drug discovery
Chen, H., Engkvist, O., Wang, Y., Olivecrona, M., & Blaschke, T. (2018) · Drug Discovery Today · doi:10.1016/j.drudis.2018.01.039 · Cited by 1,770 · see citing works
Algorithmic Learning in a Random World
(2005) · doi:10.1007/b106715 · Cited by 1,119 · see citing works
Review on Methods to Fix Number of Hidden Neurons in Neural Networks
Sheela, K. G., & Deepa, S. N. (2013) · Mathematical Problems in Engineering · doi:10.1155/2013/425740 · Cited by 956 · see citing works
K DEEP : Protein–Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks
Jiménez-Luna, J., Škalič, M., Martínez-Rosell, G., & Fabritiis, G. D. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.7b00650 · Cited by 923 · see citing works
Fourches, D., Muratov, E., & Tropsha, A. (2010) · Journal of Chemical Information and Modeling · doi:10.1021/ci100176x · Cited by 832 · see citing works
ChEMBL web services: streamlining access to drug discovery data and utilities
Davies, M., Nowotka, M., Papadatos, G., Dedman, N., Gaulton, A., Atkinson, F. L., Bellis, L. J., & Overington, J. P. (2015) · Nucleic Acids Research · doi:10.1093/nar/gkv352 · Cited by 765 · see citing works
A tutorial on conformal prediction
Shafer, G., & Vovk, V. (2007) · arXiv · doi:10.48550/arxiv.0706.3188 · Cited by 655 · see citing works
Lenselink, E. B., ten Dijke, N., Bongers, B., Papadatos, G., van Vlijmen, H. W. T., Kowalczyk, W., IJzerman, A. P., & van Westen, G. J. P. (2017) · Journal of Cheminformatics · doi:10.1186/s13321-017-0232-0 · Cited by 334 · see citing works
Comparability of Mixed IC50 Data – A Statistical Analysis
Kalliokoski, T., Krämer, C., Vulpetti, A., & Gedeck, P. (2013) · PLoS ONE · doi:10.1371/journal.pone.0061007 · Cited by 317 · see citing works
Koutsoukas, A., Monaghan, K. J., Li, X., & Huan, J. (2017) · Journal of Cheminformatics · doi:10.1186/s13321-017-0226-y · Cited by 315 · see citing works
Comparing structural fingerprints using a literature-based similarity benchmark
O’Boyle, N. M., & Sayle, R. A. (2016) · Journal of Cheminformatics · doi:10.1186/s13321-016-0148-0 · Cited by 253 · see citing works
Goh, G. B., Siegel, C., Vishnu, A., Hodas, N. O., & Baker, N. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1706.06689 · Cited by 233 · see citing works
Repurposing High-Throughput Image Assays Enables Biological Activity Prediction for Drug Discovery
Simm, J., Klambauer, G., Arany, Á., Steijaert, M. N., Wegner, J. K., Gustin, E. M., Chupakhin, V., Chong, Y. T., Vialard, J. E., Buijnsters, P. J. J. A., et al. (2018) · Cell chemical biology · doi:10.1016/j.chembiol.2018.01.015 · Cited by 228 · see citing works
Gaussian Processes: A Method for Automatic QSAR Modeling of ADME Properties
Obrezanova, O., Cśanyi, G., Gola, J. M. R., & Segall, M. (2007) · Journal of Chemical Information and Modeling · doi:10.1021/ci7000633 · Cited by 214 · see citing works
Norinder, U., Carlsson, L., Boyer, S. K., & Eklund, M. (2014) · Journal of Chemical Information and Modeling · doi:10.1021/ci5001168 · Cited by 208 · see citing works
Vovk, V. (2013) · Annals of Mathematics and Artificial Intelligence · doi:10.1007/s10472-013-9368-4 · Cited by 174 · see citing works
Bosc, N., Atkinson, F. L., Félix, E., Gaulton, A., Hersey, A., & Leach, A. R. (2019) · Journal of Cheminformatics · doi:10.1186/s13321-018-0325-4 · Cited by 167 · see citing works
Zhang, Y., & Lee, A. A. (2019) · Chemical Science · doi:10.1039/c9sc00616h · Cited by 156 · see citing works
Using Random Forest To Model the Domain Applicability of Another Random Forest Model
Sheridan, R. P. (2013) · Journal of Chemical Information and Modeling · doi:10.1021/ci400482e · Cited by 127 · see citing works
Three Useful Dimensions for Domain Applicability in QSAR Models Using Random Forest
Sheridan, R. P. (2012) · Journal of Chemical Information and Modeling · doi:10.1021/ci300004n · Cited by 125 · see citing works
Assessing the Scaffold Diversity of Screening Libraries
Krier, M., Bret, G., & Rognan, D. (2006) · Journal of Chemical Information and Modeling · doi:10.1021/ci050352v · Cited by 122 · see citing works
Improved large-scale prediction of growth inhibition patterns using the NCI60 cancer cell line panel
Cortés‐Ciriano, I., van Westen, G. J. P., Bouvier, G., Nilges, M., Overington, J. P., Bender, A., & Malliavin, T. E. (2015) · Bioinformatics · doi:10.1093/bioinformatics/btv529 · Cited by 121 · see citing works
Self-citationAssociation of Omics Features with Histopathology Patterns in Lung Adenocarcinoma
Yu, K., Berry, G. J., Rubin, D. L., Ré, C., Altman, R. B., & Snyder, M. P. (2017) · Cell Systems · doi:10.1016/j.cels.2017.10.014 · Cited by 119 · see citing works
Hofmarcher, M., Rumetshofer, E., Clevert, D.-A., Hochreiter, S., & Klambauer, G. (2019) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.8b00670 · Cited by 111 · see citing works
Cortés‐Ciriano, I., & Bender, A. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.8b00542 · Cited by 102 · see citing works
Self-citationQM/NN QSPR Models with Error Estimation: Vapor Pressure and LogP
Beck, B., Breindl, A., & Clark, T. (2000) · Journal of Chemical Information and Computer Sciences · doi:10.1021/ci990131n · Cited by 87 · see citing works
Liu, R., & Wallqvist, A. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.8b00597 · Cited by 79 · see citing works
Svensson, F., Aniceto, N., Norinder, U., Cortés‐Ciriano, I., Spjuth, O., Carlsson, L., & Bender, A. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.8b00054 · Cited by 72 · see citing works
Self-citationImputation of Assay Bioactivity Data Using Deep Learning
Whitehead, T. M., Irwin, B., Hunt, P. A., Segall, M., & Conduit, G. J. (2019) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.8b00768 · Cited by 71 · see citing works
Schroeter, T. S., Schwaighofer, A., Mika, S., ter Laak, A., Suelzle, D., Ganzer, U., Heinrich, N., & Müller, K. (2007) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-007-9125-z · Cited by 70 · see citing works
Sun, J., Carlsson, L., Ahlberg, E., Norinder, U., Engkvist, O., & Chen, H. (2017) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.7b00159 · Cited by 69 · see citing works
Quantitative Structure−Activity Relationship Studies Using Gaussian Processes
Burden, F. R. (2001) · Journal of Chemical Information and Computer Sciences · doi:10.1021/ci000459c · Cited by 65 · see citing works
Improving Screening Efficiency through Iterative Screening Using Docking and Conformal Prediction
Svensson, F., Norinder, U., & Bender, A. (2017) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.6b00532 · Cited by 65 · see citing works
Self-citationSheridan, R. P. (2015) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.5b00110 · Cited by 64 · see citing works
Towards trustable machine learning
(2018) · Nature Biomedical Engineering · doi:10.1038/s41551-018-0315-x · Cited by 64 · see citing works
Zhou, P., Tian, F., Chen, X., & Shang, Z. (2008) · Biopolymers · doi:10.1002/bip.21091 · Cited by 62 · see citing works
Wainer, H., Gessaroli, M. E., & Verdi, M. (2006) · CHANCE · doi:10.1080/09332480.2006.10722771 · Cited by 51 · see citing works
Proteochemometric modeling in a Bayesian framework
Cortés‐Ciriano, I., van Westen, G. J. P., Lenselink, E. B., Murrell, D. S., Bender, A., & Malliavin, T. E. (2014) · Journal of Cheminformatics · doi:10.1186/1758-2946-6-35 · Cited by 45 · see citing works
Self-citationEfficient iterative virtual screening with Apache Spark and conformal prediction
Ahmed, L., Georgiev, V., Capuccini, M., et al. (2018) · Journal of Cheminformatics · doi:10.1186/s13321-018-0265-z · Cited by 41 · see citing works
Wood, D., Carlsson, L., Eklund, M., et al. (2013) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-013-9639-5 · Cited by 40 · see citing works
Discovering Highly Potent Molecules from an Initial Set of Inactives Using Iterative Screening
Cortés‐Ciriano, I., Firth, N. C., Bender, A., & Watson, O. P. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.8b00376 · Cited by 39 · see citing works
Self-citationCortés‐Ciriano, I., & Bender, A. (2015) · ChemMedChem · doi:10.1002/cmdc.201500424 · Cited by 32 · see citing works
Self-citationInterpretable regression trees using conformal prediction
Johansson, U., Linusson, H., Löfström, T., & Boström, H. (2017) · Expert Systems with Applications · doi:10.1016/j.eswa.2017.12.041 · Cited by 32 · see citing works
A Distance-Based Boolean Applicability Domain for Classification of High Throughput Screening Data
Berenger, F., & Yamanishi, Y. (2018) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.8b00499 · Cited by 28 · see citing works
The challenges of making decisions using uncertain data
Segall, M., & Champness, E. J. (2015) · Journal of Computer-Aided Molecular Design · doi:10.1007/s10822-015-9855-2 · Cited by 22 · see citing works
Exploring loss function topology with cyclical learning ratesPreprint
Smith, L. N., & Topin, N. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1702.04283 · Cited by 17 · see citing works
Want Drugs? Use PythonPreprint
Nowotka, M., Papadatos, G., Davies, M., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1607.00378 · Cited by 4 · see citing works
Cortés‐Ciriano, I., & Bender, A. (2018) · arXiv (Cornell University) · Cited by 3 · see citing works
Self-citation
Flagged first, then by citations
Research works printed but not linked: 2
- [63](63) Cortes-Ciriano, I.; Bender, A. Deep Confidence: A Computationally Efficient Framework for Calculating Reliable Errors for Deep Neural Networks. J. Chem. Inf. Model. 2018 , In press , 1809.09060. -
- [66](66) Cortés-Ciriano, I.; van Westen, G. J. P.; Bouvier, G.; Nilges, M.; Overington, J. P.; Bender, A.; Malliavin, T. E. Improved Large-Scale Prediction of Growth Inhibition Patterns on the NCI60 Cancer Cell-Line Panel. Bioinformatics 2016 , 32 (1), 85-95. -
Cited by
Sorted by citations
Bioactivity descriptors for uncharacterized chemical compounds
Bertoni, M., Duran-Frigola, M., Badia-i-Mompel, P., Pauls, E., Orozco-Ruiz, M., Guitart-Pla, O., Alcalde, V., Diaz, V. M., Berenguer-Llergo, A., Brun-Heath, I., et al. (2021) · Nature Communications · doi:10.1038/s41467-021-24150-4 · Cited by 92 · see citing works
Uncertainty-aware deep learning in healthcare: A scoping review
Loftus, T. J., Shickel, B., Ruppert, M. M., Balch, J. A., Ozrazgat-Baslanti, T., Tighe, P. J., Efron, P. A., Hogan, W. R., Rashidi, P., Upchurch, G. R., et al. (2022) · PLOS Digital Health · doi:10.1371/journal.pdig.0000085 · Cited by 73 · see citing works
- Authors
- Cortés-Ciriano, I., & Bender, A.
- Affiliations
- Centre for Molecular Informatics, Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, United Kingdom
- University of Cambridge (GB)
- Venue
- Journal of Chemical Information and Modeling
- Publisher
- American Chemical Society (ACS)
- Published
- 2019-06-17
- Volume
- 59
- Issue
- 7
- Pages
- 3330-3339
- ISSN
- 1549-9596 (ISSN-L)
- 1549-960X
- OpenAlex
- W2938282337
- Cited by
- 51
- References
- 62
- Citations by year
- 2026: 2 · 2025: 2 · 2024: 11 · 2023: 6 · 2022: 4 · 2021: 12 · 2020: 9 · 2019: 5
- FWCI
- 5.11
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- Journal of Chemical Information and Modeling (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- Funders
- H2020 Marie Sklodowska-Curie Actions (703543)
- Topics
- Physical Sciences › Computer Science › Computational Theory and Mathematics › Computational Drug Discovery Methods
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
- Keywords
- conformal prediction, prediction error estimation, deep neural network, drug discovery, bioactivity prediction, confidence intervals, virtual screening, ensemble forecasting, random forest comparison, precision medicine
- MeSH terms
- Machine Learning, Neural Networks, Computer, Drug Discovery
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
