The causes of most birth defects are still poorly understood. Identification of specific cause is often challenging as there could be one single factor contributing to the defect or a mix of multiple factors. To further characterize the associations between preclinical compounds and their potentials to induce birth defects, a group of American data scientists tried to construct a knowledge graph integrating datasets on birth-defect associations noted in published work for synergistic discoveries. The data included studies specifically on genetic associations, drug- and preclinical-compound induced gene expression changes in different areas. The effort of this work is hoped to offer hypotheses about which drug or compound could be involved in the induction of birth defect.  The production of knowledge graph with the assistance of semi-supervised machine learning model helped to identify >500 birth-defect/gene/drug cliques for the explanation of molecular mechanism for drug induced birth defects. To further utilize the experience from this project, the team plans to adopt the same approach for other projects investigating associations between genes, drugs, and diseases.

Reference:

Evangelista JE, et al. Commun Med (Lond). 2023 Jul 17;3(1):98.