Generalising uncertainty improves accuracy and safety of deep learning analytics applied to oncology
Open accessPublished version of a preprint
Abstract
Uncertainty estimation is crucial for understanding the reliability of deep learning (DL) predictions, and critical for deploying DL in the clinic. Differences between training and production datasets can lead to incorrect predictions with underestimated uncertainty. To investigate this pitfall, we benchmarked one pointwise and three approximate Bayesian DL models for predicting cancer of unknown primary, using three RNA-seq datasets with 10,968 samples across 57 cancer types. Our results highlight that simple and scalable Bayesian DL significantly improves the generalisation of uncertainty estimation. Moreover, we designed a prototypical metric—the area between development and production curve (ADP), which evaluates the accuracy loss when deploying models from development to production. Using ADP, we demonstrate that Bayesian DL improves accuracy under data distributional shifts when utilising ‘uncertainty thresholding’. In summary, Bayesian DL is a promising approach for generalising uncertainty, improving performance, transparency, and safety of DL models for deployment in the real world.
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.
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
Highly accurate protein structure prediction with AlphaFold
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., et al. (2021) · Nature · doi:10.1038/s41586-021-03819-2 · Cited by 43,764 · 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
On calibration of modern neural networks
Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. (2017) · arXiv · doi:10.48550/arxiv.1706.04599 · Cited by 3,435 · see citing works
A review of uncertainty quantification in deep learning: Techniques, applications and challenges
Abdar, M., Pourpanah, F., Hussain, S., Rezazadegan, D., Liu, L., Ghavamzadeh, M., Fieguth, P., Cao, X., Khosravi, A., Acharya, U. R., et al. (2021) · Information Fusion · doi:10.1016/j.inffus.2021.05.008 · Cited by 2,720 · cited 1× · see citing works
“Epistemic and aleatoric uncertainties present the potential for further insights, including whether a data point’s predictive uncertainty will reduce with either more examples or by an altered model design (epistemic uncertainty), or more features (aleatoric uncertainty) 51–54 .”
BackgroundHoadley, K. A., Yau, C., Hinoue, T., Wolf, D. M., Lazar, A. J., Drill, E., Shen, R., Taylor, A. M., Cherniack, A. D., Thorsson, V., et al. (2018) · Cell · doi:10.1016/j.cell.2018.03.022 · Cited by 2,546 · cited 1× · see citing works
“The test data were OOD (representing ‘production’), providing a platform for benchmarking resilience to overconfidence, and included TCGA metastatic samples (n=392; [32]), Met500 metastatic samples (n=479; [33]), and a combination of primary and metastatic samples from our own independent internal custom dataset, i.”
BackgroundAleatoric and epistemic uncertainty in machine learning: an introduction to concepts and methods
Hüllermeier, E., & Waegeman, W. (2021) · Machine Learning · doi:10.1007/s10994-021-05946-3 · Cited by 1,834 · see citing works
Whole-genome landscape of pancreatic neuroendocrine tumours
Initiative, A. P. C. G., Scarpa, A., Chang, D. K., Nones, K., Corbo, V., Patch, A.-M., Bailey, P., Lawlor, R. T., Johns, A. L., Miller, D. K., et al. (2017) · Nature · doi:10.1038/nature21063 · Cited by 999 · cited 1× · see citing works
“Three independent datasets were used: our own independent Internal Custom Dataset, ICD 34–42 , TCGA 31 , and Met500 33 .”
MethodSelf-citationCan You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift
Ovadia, Y., Fertig, E., Ren, J., Nado, Z., Sculley, D., Nowozin, S., Dillon, J. V., Lakshminarayanan, B., & Snoek, J. (2019) · arXiv · doi:10.48550/arxiv.1906.02530 · Cited by 989 · see citing works
A Python library for probabilistic analysis of single-cell omics data
Gayoso, A., Lopez, R., Xing, G., Boyeau, P., Valiollah Pour Amiri, V., Hong, J., Wu, K., Jayasuriya, M., Mehlman, E., Langevin, M., et al. (2022) · Nature Biotechnology · doi:10.1038/s41587-021-01206-w · Cited by 789 · cited 1× · see citing works
“The proposed applications of DL utilise a range of data modalities, including MRI scans [5], CT scans [6], histopathology slides [7], genomics [8], transcriptomics [9], [10], and most recently, integrated approaches with various data types [11], [12].”
MethodA Survey of Uncertainty in Deep Neural Networks
Gawlikowski, J., Tassi, C. R. N., Ali, M., Lee, J., Humt, M., Feng, J., Kruspe, A., Triebel, R., Jung, P., Roscher, R., et al. (2021) · arXiv · doi:10.48550/arxiv.2107.03342 · Cited by 232 · see citing works
Grewal, J. K., Tessier-Cloutier, B., Jones, M., Gakkhar, S., Ma, Y., Moore, R., Mungall, A. J., Zhao, Y., Taylor, M. D., Gelmon, K., et al. (2019) · JAMA Network Open · doi:10.1001/jamanetworkopen.2019.2597 · Cited by 90 · see citing works
Divate, M., Tyagi, A., Richard, D. J., Prasad, P. A., Gowda, H., & Nagaraj, S. H. (2022) · Cancers · doi:10.3390/cancers14051185 · Cited by 55 · see citing works
Flagged first, then by citations
- Authors
- MacDonald, S., Foley, H., Yap, M., Johnston, R. L., Steven, K., Koufariotis, L. T., Sharma, S., Wood, S., Addala, V., Pearson, J. V., Roosta, F., Waddell, N., Kondrashova, O., & Trzaskowski, M.
- Affiliations
- The University of Queensland (AU)
- QIMR Berghofer Medical Research Institute (AU)
- Corresponding author
- Kondrashova, Olga
- Trzaskowski, Maciej
- ORCID
- Venue
- Scientific Reports
- Published
- 2023-05-06
- Volume
- 13
- Issue
- 1
- Pages
- 7395-7395
- ISSN
- 2045-2322 (ISSN-L)
- PMC
- PMC10164181
- OpenAlex
- W4372354552
- Cited by
- 12
- References
- 12
- Citations by year
- 2026: 5 · 2025: 3 · 2024: 2 · 2023: 2
- FWCI
- 1.13
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- Scientific Reports (publishedVersion, cc-by)
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
- Research Square (acceptedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Cooperative Research Centres, Australian Government Department of Industry (CRCPFIVE000176)
- Australian Research Council Industrial Transformation Training Centre for Information Resilience (IC200100022)
- National Health and Medical Research Council (APP1139071)
- NHMRC Emerging Leader 1 Investigator Grant (APP2008631)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
- Health Sciences › Medicine › Oncology › Cancer Diagnosis and Treatment
- Keywords
- uncertainty estimation, cancer of unknown primary, RNA-seq, model generalization
- MeSH terms
- Deep Learning, Bayes Theorem, Medical Oncology, Reproducibility of Results, Uncertainty
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
