Representation Learning of Human Disease Mechanisms for a Foundation Model in Rare and Common Diseases
Ravandi, B.; Mowrey, W. R.; Chatterjee, A.; Haddadi, P.; Abdelmessih, M.; Ding, W.; Lambden, S.; Ughetto, M.; Barrett, I.; Diethe, T.; Del Angel, G.; Eliassi-Rad, T.; Ricchiuto, P.
Show abstract
The limited amount of data available renders it challenging to characterize which biological processes are relevant to a rare disease. Hence, there is a need to leverage the knowledge of disease pathogenesis and treatment from the wider disease landscape to understand rare disease mechanisms. Furthermore, it is well understood that rare disease discoveries can inform the our knowledge of common diseases. In this paper, we introduce Dis2Vec (Disease to Vector), a new representation learning method for characterizing diseases with a focus on learning the underlying biological mechanisms, which is a step toward developing a foundation model for disease-association learning. Dis2Vec is trained on human genetic evidence and observed symptoms, and then evaluated through cross-modal transfer-learning scenarios based on a proposed drug association learning benchmark with drug targets (positive controls) and Orphanet Rare Disease Ontology (negative controls). Finally, we argue that clustering diseases in the Dis2Vec space, which captures biological mechanisms instead of drug-repurposing information, could increase the efficiency of translational research in rare and common diseases, and ultimately improve treatment strategies for patients.
Matching journals
The top 8 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- Learning interpretable cellular and gene signature embeddings from single-cell transcriptomic data 96%
- Deep representation learning for clustering longitudinal survival data from electronic health records 95%
- Projecting genetic associations through gene expression patterns highlights disease etiology and drug mechanisms 95%
Similar papers in this journal
- scELMo: Embeddings from Language Models are Good Learners for Single-cell Data Analysis 95%
- Bi-level Graph Learning Unveils Prognosis-Relevant Tumor Microenvironment Patterns in Breast Multiplexed Digital Pathology 95%
- scTenifoldNet: a machine learning workflow for constructing and comparing transcriptome-wide gene regulatory networks from single-cell data 94%
Similar papers in this journal
- Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases 97%
- FedWeight: Mitigating Covariate Shift of Federated Learning on Electronic Health Records Data through Patients Re-weighting 95%
- Clinical Knowledge Extraction via Sparse Embedding Regression (KESER) with Multi-Center Large Scale Electronic Health Record Data 95%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.