In this post, we share insights from Serokell AI experts on their investigation into drug-disease interactions. Specifically, they explored whether a drug has a positive, negative, or neutral effect on the treatment of a particular disease.
Serokell has collaborated with Neo7Bioscience, a molecular technology company, and Elsevier, an information and analytics firm that facilitates medical and biological research. With the data licensed from Elsevier, our specialists developed ML models that predict interactions between small molecules and diseases.
Drug-disease interaction prediction and biological sequence embedding
The task is related to big data analysis and consisted of assisting Elsevier with handling a large dataset derived from tens of thousands of research papers, which was condensed into a large graph. While not very large compared to datasets used in other ML fields like NLP and computer vision, this dataset contains a diverse array of biological entities such as diseases, proteins, and small molecules, along with various types of connections between them, notably clinical trials and regulations.
We focused on two major tasks: drug-disease interaction prediction and biological sequence embedding.
- Drug-disease…
