CoChat
Sign Up
Cited by 111Bioinformatics and Genomic NetworksComputational Drug Discovery MethodsMachine Learning in Healthcare

The DOI matches its Crossref record; 75 of the 76 references it cites were found in a registry or an index.

Discovery of drug-omics associations in type 2 diabetes with generative deep-learning models

Open access
Allesøe, Rosa Lundbye, Lundgaard, Agnete Troen, Hernández Medina, Ricardo, Aguayo-Orozco, Alejandro, Johansen, Joachim, Nissen, Jakob Nybo, Brorsson, Caroline, Mazzoni, Gianluca, Niu, Lili, Biel, Jorge Hernansanz, Leal Rodríguez, Cristina, Brasas, Valentas, Webel, Henry, Benros, Michael Eriksen, Pedersen, Anders Gorm, Chmura, Piotr Jaroslaw, Jacobsen, Ulrik Plesner, Mari, Andrea, Koivula, Robert, Mahajan, Anubha, Vinuela, Ana, Tajes, Juan Fernandez, Sharma, Sapna, Haid, Mark, Hong, Mun-Gwan, Musholt, Petra B., De Masi, Federico, Vogt, Josef, Pedersen, Helle Krogh, Gudmundsdottir, Valborg, Jones, Angus, Kennedy, Gwen, Bell, Jimmy, Thomas, E. Louise, Frost, Gary, Thomsen, Henrik, Hansen, Elizaveta, Hansen, Tue Haldor, Vestergaard, Henrik, Muilwijk, Mirthe, Blom, Marieke T., ‘t Hart, Leen M., Pattou, Francois, Raverdy, Violeta, Brage, Soren, Kokkola, Tarja, Heggie, Alison, McEvoy, Donna, Mourby, Miranda, Kaye, Jane, Hattersley, Andrew, McDonald, Timothy, Ridderstråle, Martin, Walker, Mark, Forgie, Ian, Giordano, Giuseppe N., Pavo, Imre, Ruetten, Hartmut, Pedersen, Oluf, Hansen, Torben, Dermitzakis, Emmanouil, Franks, Paul W., Schwenk, Jochen M., Adamski, Jerzy, McCarthy, Mark I., Pearson, Ewan, Banasik, Karina, Rasmussen, Simon, Brunak, Søren, IMI DIRECT Consortium, Froguel, Philippe, Thomas, Cecilia Engel, Haussler, Ragna, Beulens, Joline, Rutters, Femke, Nijpels, Giel, van Oort, Sabine, Groeneveld, Lenka, Elders, Petra, Giorgino, Toni, Rodriquez, Marianne, Nice, Rachel, Perry, Mandy, Bianzano, Susanna, Graefe-Mody, Ulrike, Hennige, Anita, Grempler, Rolf, Baum, Patrick, Stærfeldt, Hans-Henrik, Shah, Nisha, Teare, Harriet, Ehrhardt, Beate, Tillner, Joachim, Dings, Christiane, Lehr, Thorsten, Scherer, Nina, Sihinevich, Iryna, Cabrelli, Louise, Loftus, Heather, Bizzotto, Roberto, Tura, Andrea, Dekkers, Koen, van Leeuwen, Nienke, Groop, Leif, Slieker, Roderick, Ramisch, Anna, Jennison, Christopher, McVittie, Ian, Frau, Francesca, Steckel-Hamann, Birgit, Adragni, Kofi, Thomas, Melissa, Pasdar, Naeimeh Atabaki, Fitipaldi, Hugo, Kurbasic, Azra, Mutie, Pascal, Pomares-Millan, Hugo, Bonnefond, Amelie, Canouil, Mickael, Caiazzo, Robert, Verkindt, Helene, Holl, Reinhard, Kuulasmaa, Teemu, Deshmukh, Harshal, Cederberg, Henna, Laakso, Markku, Vangipurapu, Jagadish, Dale, Matilda, Thorand, Barbara, Nicolay, Claudia, Fritsche, Andreas, Hill, Anita, Hudson, Michelle, Thorne, Claire, Allin, Kristine, Arumugam, Manimozhiyan, Jonsson, Anna, Engelbrechtsen, Line, Forman, Annemette, Dutta, Avirup, Sondertoft, Nadja, Fan, Yong, Gough, Stephen, Robertson, Neil, McRobert, Nicky, Wesolowska-Andersen, Agata, Brown, Andrew, Davtian, David, Dawed, Adem, Donnelly, Louise, Palmer, Colin, White, Margaret, Ferrer, Jorge, Whitcher, Brandon, Artati, Anna, Prehn, Cornelia, Adam, Jonathan, Grallert, Harald, Gupta, Ramneek, Sackett, Peter Wad, Nilsson, Birgitte, Tsirigos, Konstantinos, Eriksen, Rebeca, Jablonka, Bernd, Uhlen, Mathias, Gassenhuber, Johann, Baltauss, Tania, de Preville, Nathalie, Klintenberg, Maria, Abdalla, Moustafa·2023·Nature Biotechnology
DOI 10.1038/s41587-022-01520-xPMC PMC10017515
Open this paper in CoChat — free accountOpen PDF

Abstract

The application of multiple omics technologies in biomedical cohorts has the potential to reveal patient-level disease characteristics and individualized response to treatment. However, the scale and heterogeneous nature of multi-modal data makes integration and inference a non-trivial task. We developed a deep-learning-based framework, multi-omics variational autoencoders (MOVE), to integrate such data and applied it to a cohort of 789 people with newly diagnosed type 2 diabetes with deep multi-omics phenotyping from the DIRECT consortium. Using in silico perturbations, we identified drug–omics associations across the multi-modal datasets for the 20 most prevalent drugs given to people with type 2 diabetes with substantially higher sensitivity than univariate statistical tests. From these, we among others, identified novel associations between metformin and the gut microbiota as well as opposite molecular responses for the two statins, simvastatin and atorvastatin. We used the associations to quantify drug–drug similarities, assess the degree of polypharmacy and conclude that drug effects are distributed across the multi-omics modalities.

Overview

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.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41587-022-01520-x

Cites

75 of 75 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Integrating multi-omics data with a VAE. a, Principle of integration and analysis approach using MOVE. Individual-level non-omics and multi-omics data were used as input to a VAE. The optimal network hyperparameters were estimated from the summed test set error across all individuals in the test (test likelihood), training reconstruction accuracy, and model stability. Significant drug–omics associations were identified by perturbing drug status from no (0) to yes (1) for all individuals that were not already administered the drug. b, UMAP representation of the latent representation from the 789 people with newly diagnosed T2D. Individuals were colored according to their z-scaled Matsuda index from low (blue), average (yellow), and high (red). c, Overlap in significant drug–omics associations between standard t-test (two-sided, Benjamini–Hochberg FDR < 0.01) on the input data, MOVE t-test (multi-stage Bonferroni-corrected, P adjust < 0.05) and MOVE Bayes approaches (FDR Bayes < 0.05). The different methods of multiple testing correction corresponded to FDR of 0.05 on the ground-truth dataset. The overlap between MOVE t-test and MOVE Bayes was used for further analysis (n = 573). d, The number of significant associations found between drugs and features in the multi-omics datasets using MOVE t-test and MOVE Bayes (purple), t-test (green) or ANOVA (orange). See c for information on the tests. e, Fraction of features in the multi-omics datasets that was found by MOVE to be significantly associated with at least one drug (n = 20). The lower and upper hinges correspond to the first and third quartiles. The upper and lower whiskers extend from the hinge to the highest and lowest values, respectively, but no further than 1.5× interquartile range from the hinge. Data beyond the ends of whiskers are outliers and are plotted individually. Source data
Fig. 1 Integrating multi-omics data with a VAE. a, Principle of integration and analysis approach using MOVE. Individual-level non-omics and multi-omics data were used as input to a VAE. The optimal network hyperparameters were estimated from the summed test set error across all individuals in the test (test likelihood), training reconstruction accuracy, and model stability. Significant drug–omics associations were identified by perturbing drug status from no (0) to yes (1) for all individuals that were not already administered the drug. b, UMAP representation of the latent representation from the 789 people with newly diagnosed T2D. Individuals were colored according to their z-scaled Matsuda index from low (blue), average (yellow), and high (red). c, Overlap in significant drug–omics associations between standard t-test (two-sided, Benjamini–Hochberg FDR < 0.01) on the input data, MOVE t-test (multi-stage Bonferroni-corrected, P adjust < 0.05) and MOVE Bayes approaches (FDR Bayes < 0.05). The different methods of multiple testing correction corresponded to FDR of 0.05 on the ground-truth dataset. The overlap between MOVE t-test and MOVE Bayes was used for further analysis (n = 573). d, The number of significant associations found between drugs and features in the multi-omics datasets using MOVE t-test and MOVE Bayes (purple), t-test (green) or ANOVA (orange). See c for information on the tests. e, Fraction of features in the multi-omics datasets that was found by MOVE to be significantly associated with at least one drug (n = 20). The lower and upper hinges correspond to the first and third quartiles. The upper and lower whiskers extend from the hinge to the highest and lowest values, respectively, but no further than 1.5× interquartile range from the hinge. Data beyond the ends of whiskers are outliers and are plotted individually. Source dataCC BY · PubMed Central
Fig. 2 Significant associations between drugs, clinical, and multi-omics features. a, Significant associations between drugs and clinical features. Effects are given as effect size (z-scaled units) from negative (blue) to positive (red). Significant associations identified by both MOVE t-test and MOVE Bayes are indicated using a star. Features (y-axis) and drugs (x-axis) are clustered using hierarchical clustering on the basis of Euclidean distances. b, As in a but showing per individual-level associations of metformin to multi-omics features demonstrating that associations are highly stable across individuals. Features (y-axis) and newly diagnosed T2D individuals (x-axis). Source data
Fig. 2 Significant associations between drugs, clinical, and multi-omics features. a, Significant associations between drugs and clinical features. Effects are given as effect size (z-scaled units) from negative (blue) to positive (red). Significant associations identified by both MOVE t-test and MOVE Bayes are indicated using a star. Features (y-axis) and drugs (x-axis) are clustered using hierarchical clustering on the basis of Euclidean distances. b, As in a but showing per individual-level associations of metformin to multi-omics features demonstrating that associations are highly stable across individuals. Features (y-axis) and newly diagnosed T2D individuals (x-axis). Source dataCC BY · PubMed Central
Fig. 3 Drug associations with metagenomics species and drug–drug similarities. a, Display of effect sizes (z-scaled units) for (outer to inner) metformin, simvastatin, atorvastatin, omeprazole, lansoprazole, paracetamol, and codeine. Only significant associations to any of the drugs are shown and effect size is visualized as brown (negative), gray (none), and green (positive). Selected omics features are indicated. The Gene Ontologies element represents significantly over-represented Gene Ontology terms using transcriptomics (hypergeometric test, FDR < 0.05) (green). The innermost ring indicates SHAP importance for the individual features in the encoding from input data to the latent representation. b, Effect size (z-scaled units) (x-axis) of the human gut metagenomics species that were significantly associated with metformin (orange) or omeprazole (teal). c, Drug–drug similarities by comparing drug-response profiles across the multi-omics datasets. Cosine similarity indicated from no similarity (blue) to identical profiles (red). d, Average effect (z-score) of drugs for the omics datasets. All 20 drugs are shown, however, only metformin (red), omeprazole (purple), atorvastatin (green), and simvastatin (blue) are indicated. All other drugs are colored gray without a text label. e, Distribution of multi-omics ranks for the different drugs. The ranks are determined as a number between 1–20 (drugs) on the basis of the average effect size from d. The boxes are colored according to number of individuals taking a particular drug from 0 (white) to 323 (purple). There was no correlation between rank scores and number of individuals taking a drug (PCC = 0.14). The lower and upper hinges correspond to the first and third quartiles. The upper and lower whiskers extend from the hinge to the highest and lowest values, respectively, but no further than 1.5× interquartile range from the hinge. Data beyond the ends of whiskers are outliers and are plotted individually. Source data
Fig. 3 Drug associations with metagenomics species and drug–drug similarities. a, Display of effect sizes (z-scaled units) for (outer to inner) metformin, simvastatin, atorvastatin, omeprazole, lansoprazole, paracetamol, and codeine. Only significant associations to any of the drugs are shown and effect size is visualized as brown (negative), gray (none), and green (positive). Selected omics features are indicated. The Gene Ontologies element represents significantly over-represented Gene Ontology terms using transcriptomics (hypergeometric test, FDR < 0.05) (green). The innermost ring indicates SHAP importance for the individual features in the encoding from input data to the latent representation. b, Effect size (z-scaled units) (x-axis) of the human gut metagenomics species that were significantly associated with metformin (orange) or omeprazole (teal). c, Drug–drug similarities by comparing drug-response profiles across the multi-omics datasets. Cosine similarity indicated from no similarity (blue) to identical profiles (red). d, Average effect (z-score) of drugs for the omics datasets. All 20 drugs are shown, however, only metformin (red), omeprazole (purple), atorvastatin (green), and simvastatin (blue) are indicated. All other drugs are colored gray without a text label. e, Distribution of multi-omics ranks for the different drugs. The ranks are determined as a number between 1–20 (drugs) on the basis of the average effect size from d. The boxes are colored according to number of individuals taking a particular drug from 0 (white) to 323 (purple). There was no correlation between rank scores and number of individuals taking a drug (PCC = 0.14). The lower and upper hinges correspond to the first and third quartiles. The upper and lower whiskers extend from the hinge to the highest and lowest values, respectively, but no further than 1.5× interquartile range from the hinge. Data beyond the ends of whiskers are outliers and are plotted individually. Source dataCC BY · PubMed Central

Verification

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.

A few free checks a day · papers already verified open instantly · results are public

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?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 76 references

Research works printed but not linked: 5

  1. Mattei, P.-A. & Frellsen, J. MIWAE: deep generative modelling and imputation of incomplete data. In Proceedings of the 36th International Conference on Machine Learning 4413–4423 (PMLR, 2019).
  2. Lopez, R., Boyeau, P., Yosef, N., Jordan, M. I. & Regier, J. Decision-making with auto-encoding variational Bayes. In Proceedings of the 34th International Conference on Neural Information Processing Systems 5081–5092 (Curran Associates Inc., 2020).
  3. Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. in Advances in Neural Information Processing Systems 30 (eds Guyon, I. et al.) 4765–4774 (Curran Associates, 2017).
  4. Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I. & Salakhutdinov, R. Dropout: a simple way to prevent neural networks from overfitting. J. Mach. Learn. Res. 15, 1929–1958 (2014).
  5. Maas, A. L., Hannun, A. Y. & Ng, A. Y. Rectifier nonlinearities improve neural network acoustic models. In Proc. 30th International Conference on Machine Learning (eds Dasgupta, S. & McAllester, D.) (JMLR, 2013).

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Allesøe, R. L., Lundgaard, A. T., Hernández Medina, R., Aguayo-Orozco, A., Johansen, J., Nissen, J. N., Brorsson, C., Mazzoni, G., Niu, L., Biel, J. H., Leal Rodríguez, C., Brasas, V., Webel, H., Benros, M. E., Pedersen, A. G., Chmura, P. J., Jacobsen, U. P., Mari, A., Koivula, R., Mahajan, A., Vinuela, A., Tajes, J. F., Sharma, S., Haid, M., Hong, M.-G., Musholt, P. B., De Masi, F., Vogt, J., Pedersen, H. K., Gudmundsdottir, V., Jones, A., Kennedy, G., Bell, J., Thomas, E. L., Frost, G., Thomsen, H., Hansen, E., Hansen, T. H., Vestergaard, H., Muilwijk, M., Blom, M. T., ‘t Hart, L. M., Pattou, F., Raverdy, V., Brage, S., Kokkola, T., Heggie, A., McEvoy, D., Mourby, M., Kaye, J., Hattersley, A., McDonald, T., Ridderstråle, M., Walker, M., Forgie, I., Giordano, G. N., Pavo, I., Ruetten, H., Pedersen, O., Hansen, T., Dermitzakis, E., Franks, P. W., Schwenk, J. M., Adamski, J., McCarthy, M. I., Pearson, E., Banasik, K., Rasmussen, S., Brunak, S., IMI DIRECT Consortium, Froguel, P., Thomas, C. E., Haussler, R., Beulens, J., Rutters, F., Nijpels, G., van Oort, S., Groeneveld, L., Elders, P., Giorgino, T., Rodriquez, M., Nice, R., Perry, M., Bianzano, S., Graefe-Mody, U., Hennige, A., Grempler, R., Baum, P., Stærfeldt, H.-H., Shah, N., Teare, H., Ehrhardt, B., Tillner, J., Dings, C., Lehr, T., Scherer, N., Sihinevich, I., Cabrelli, L., Loftus, H., Bizzotto, R., Tura, A., Dekkers, K., van Leeuwen, N., Groop, L., Slieker, R., Ramisch, A., Jennison, C., McVittie, I., Frau, F., Steckel-Hamann, B., Adragni, K., Thomas, M., Pasdar, N. A., Fitipaldi, H., Kurbasic, A., Mutie, P., Pomares-Millan, H., Bonnefond, A., Canouil, M., Caiazzo, R., Verkindt, H., Holl, R., Kuulasmaa, T., Deshmukh, H., Cederberg, H., Laakso, M., Vangipurapu, J., Dale, M., Thorand, B., Nicolay, C., Fritsche, A., Hill, A., Hudson, M., Thorne, C., Allin, K., Arumugam, M., Jonsson, A., Engelbrechtsen, L., Forman, A., Dutta, A., Sondertoft, N., Fan, Y., Gough, S., Robertson, N., McRobert, N., Wesolowska-Andersen, A., Brown, A., Davtian, D., Dawed, A., Donnelly, L., Palmer, C., White, M., Ferrer, J., Whitcher, B., Artati, A., Prehn, C., Adam, J., Grallert, H., Gupta, R., Sackett, P. W., Nilsson, B., Tsirigos, K., Eriksen, R., Jablonka, B., Uhlen, M., Gassenhuber, J., Baltauss, T., de Preville, N., Klintenberg, M., & Abdalla, M.
Affiliations
  • University of Copenhagen (DK)
  • Technical University of Denmark (DK)
  • Centre for Human Genetics (GB)
  • Helmholtz Munich (DE)
  • Science for Life Laboratory (SE)
  • Sanofi (Germany) (DE)
  • University of Exeter (GB)
  • University of Dundee (GB)
  • University of Westminster (GB)
  • Imperial College London (GB)
  • Herlev Hospital (DK)
  • Amsterdam Public Health (NL)
  • Inserm (FR)
  • University of Eastern Finland (FI)
  • Newcastle University (GB)
  • Royal Victoria Infirmary (GB)
  • University of Oxford (GB)
  • Lund University (SE)
  • Eli Lilly (United States) (US)
  • University of Geneva (CH)
  • Harvard University (US)
  • University of Ljubljana (SI)
  • Institut de Recherches Internationales Servier (France) (FR)
  • Royal Devon & Exeter NHS Foundation Trust (GB)
  • Boehringer Ingelheim (Germany) (DE)
  • University of Bath (GB)
  • Saarland University (DE)
  • Leiden University Medical Center (NL)
  • Universität Ulm (DE)
  • Eli Lilly (Germany) (DE)
  • University of Tübingen (DE)
  • Centro de Investigación Biomédica en Red Diabetes y Enfermedades Metabólicas Asociadas (ES)
Corresponding author
  • Rasmussen, Simon
  • Brunak, Søren
ORCID
Venue
Nature Biotechnology
Publisher
Springer Science and Business Media LLC
Published
2023-01-02
Volume
41
Issue
3
Pages
399-408
ISSN
  • 0733-222X (ISSN-L)
  • 1087-0156
  • 1546-1696
DOI
10.1038/s41587-022-01520-x
PMC
PMC10017515
OpenAlex
W4313455000
Cited by
111
References
76
Citations by year
2026: 25 · 2025: 35 · 2024: 34 · 2023: 17
FWCI
9.89
Type
journal article
Language
en
Update notices
correction (2023-05-02)
Open access
Licence
Locations
  • Nature Biotechnology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Discovery Research Portal (University of Dundee) (publishedVersion, cc-by)
  • Amsterdam UMC (VU Amsterdam) - Institutional Repository (publishedVersion, other-oa)
  • Research at the University of Copenhagen (University of Copenhagen) (publishedVersion, cc-by)
  • Repositori digital de la UPF (Universitat Pompeu Fabra) (acceptedVersion, cc-by)
  • WestminsterResearch (University of Westminster) (acceptedVersion, cc-by)
  • Pure Amsterdam UMC (submittedVersion, cc-by)
  • Technical University of Denmark, DTU Orbit (Technical University of Denmark, DTU) (submittedVersion, cc-by)
  • Leiden Repository (Leiden University) (submittedVersion, cc-by)
  • Apollo (University of Cambridge) (submittedVersion, cc-by)
  • CNR ExploRA (submittedVersion, cc-by)
  • Data Archiving and Networked Services (DANS) (submittedVersion, other-oa)
  • Data Archiving and Networked Services (DANS) (submittedVersion, cc-by)
  • HAL (submittedVersion)
  • LillOA (Université de Lille (University Of Lille)) (submittedVersion, cc-by)
  • Lund University Publications (Lund University) (submittedVersion)
  • mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) (submittedVersion)
  • Apollo (cc-by)
  • Apollo
Funders
  • Novo Nordisk Fonden (NNF14CC0001, NNF17OC0027594)
  • Innovative Medicines Initiative (115317)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Bioinformatics and Genomic Networks
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
Keywords
type 2 diabetes, multi-omics integration, variational autoencoder, metformin, gut microbiota, statins, simvastatin, atorvastatin, polypharmacy, drug-drug similarity
MeSH terms
Deep Learning, Algorithms, Diabetes Mellitus, Type 2, Humans

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