Multimodal representation learning for predicting molecule-disease relations

Jun Wen^{1

2}, Xiang Zhang¹, Everett Rush³, Vidul A Panickan^{1

2}, Xingyu Li¹, Tianrun Cai^{2

4}, Doudou Zhou⁵, Yuk-Lam Ho², Lauren Costa², Edmon Begoli³, Chuan Hong^{2

6}, J Michael Gaziano^{1

2

7}, Kelly Cho^{1

2

7}, Junwei Lu^{2

8}, Katherine P Liao^{1

2

8}, Marinka Zitnik^{1

9

10}, Tianxi Cai^{1

2

4}

Affiliations

¹ Department of Biomedical Informatics, Harvard Medical School, Boston, MA 02115, USA.
² VA Boston Healthcare System, Boston, MA 02130, USA.
³ Department of Energy, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA.
⁴ Mass General Brigham, Boston, MA 02130, USA.
⁵ Department of Statistics, University of California, Davis, CA 95616, USA.
⁶ Department of Biostatistics and Bioinformatics, Duke University, Durham, NC 27708, USA.
⁷ Brigham and Women's Hospital, Boston, MA 02115, USA.
⁸ Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA.
⁹ Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA.
¹⁰ Harvard Data Science Initiative, Cambridge, MA 02138, USA.

PMID: 36805623
PMCID: PMC9940625
DOI: 10.1093/bioinformatics/btad085

Multimodal representation learning for predicting molecule-disease relations

Jun Wen et al. Bioinformatics. 2023.

. 2023 Feb 3;39(2):btad085.

doi: 10.1093/bioinformatics/btad085.

Authors

Affiliations

¹ Department of Biomedical Informatics, Harvard Medical School, Boston, MA 02115, USA.
² VA Boston Healthcare System, Boston, MA 02130, USA.
³ Department of Energy, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA.
⁴ Mass General Brigham, Boston, MA 02130, USA.
⁵ Department of Statistics, University of California, Davis, CA 95616, USA.
⁶ Department of Biostatistics and Bioinformatics, Duke University, Durham, NC 27708, USA.
⁷ Brigham and Women's Hospital, Boston, MA 02115, USA.
⁸ Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA.
⁹ Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA.
¹⁰ Harvard Data Science Initiative, Cambridge, MA 02138, USA.

PMID: 36805623
PMCID: PMC9940625
DOI: 10.1093/bioinformatics/btad085

Abstract

Motivation: Predicting molecule-disease indications and side effects is important for drug development and pharmacovigilance. Comprehensively mining molecule-molecule, molecule-disease and disease-disease semantic dependencies can potentially improve prediction performance.

Methods: We introduce a Multi-Modal REpresentation Mapping Approach to Predicting molecular-disease relations (M2REMAP) by incorporating clinical semantics learned from electronic health records (EHR) of 12.6 million patients. Specifically, M2REMAP first learns a multimodal molecule representation that synthesizes chemical property and clinical semantic information by mapping molecule chemicals via a deep neural network onto the clinical semantic embedding space shared by drugs, diseases and other common clinical concepts. To infer molecule-disease relations, M2REMAP combines multimodal molecule representation and disease semantic embedding to jointly infer indications and side effects.

Results: We extensively evaluate M2REMAP on molecule indications, side effects and interactions. Results show that incorporating EHR embeddings improves performance significantly, for example, attaining an improvement over the baseline models by 23.6% in PRC-AUC on indications and 23.9% on side effects. Further, M2REMAP overcomes the limitation of existing methods and effectively predicts drugs for novel diseases and emerging pathogens.

Availability and implementation: The code is available at https://github.com/celehs/M2REMAP, and prediction results are provided at https://shiny.parse-health.org/drugs-diseases-dev/.

Supplementary information: Supplementary data are available at Bioinformatics online.

PubMed Disclaimer

Figures

**Fig. 1.**
Overview of M2REMAP. By learning clinical semantic embeddings from EHR data, M2REMAP synthesized molecule chemicals and EHR semantics to attain multimodal molecule representation combined with disease EHR semantics to jointly infer indications and side effects

**Fig. 2**
Drug–disease embedding visualization. We visualize the EHR semantic embedding of cancer and psychotropic drugs and their reported indications and side effects. (a) drugs and indications; (b) drugs and side effects

See this image and copyright information in PMC

References

1. Allegretti M. et al. (2022) Repurposing the estrogen receptor modulator raloxifene to treat SARS-COV-2 infection. Cell Death Differ., 29, 156–166. - PMC - PubMed
1. Alves V.M. et al. (2021) QSAR modeling of SARS-CoV Mpro inhibitors identifies sufugolix, cenicriviroc, proglumetacin, and other drugs as candidates for repurposing against SARS-CoV-2. Mol. Inf., 40, 2000113. - PubMed
1. Beam A.L. et al. (2019) Clinical concept embeddings learned from massive sources of multimodal medical data. In: Pacific Symposium on Biocomputing 2020, Hawaii, pp. 295–306. World Scientific. - PMC - PubMed
1. Bernstein L.R., Zhang L. (2020) Gallium maltolate has in vitro antiviral activity against SARS-CoV-2 and is a potential treatment for COVID-19. Antivir. Chem. Chemother., 28, 2040206620983780. - PMC - PubMed
1. Chandak P. et al. (2022) Building a knowledge graph to enable precision medicine. Scientific Data, 10(1), 67. - PMC - PubMed

Publication types

Actions
Actions

MeSH terms

Actions
Actions
Actions
Actions
Actions
Actions

Grants and funding

LinkOut - more resources

Full Text Sources
- Europe PubMed Central
Medical
- MedlinePlus Health Information

Save citation to file

Email citation

Add to Collections

Add to My Bibliography

Your saved search

Create a file for external citation management software

Your RSS Feed

Multimodal representation learning for predicting molecule-disease relations

Affiliations

Multimodal representation learning for predicting molecule-disease relations

Authors

Affiliations

Abstract

Figures

References

Publication types

MeSH terms

Grants and funding

LinkOut - more resources

Full Text Sources

Medical