Multimodal graph machine learning for drug target prioritization
MultiGML combines relevant knowledge and experimental data (e.g., gene expression data, microscopy images, protein sequences) in a comprehensive, structured, and unbiased manner. Technically, this is realized by a semantically harmonized knowledge graph compiled from 14 curated biological databases, resulting in approximately 400,000 relationships between proteins, drugs, and phenotypes, including ADEs.
Fraunhofer SCAI offers two versions of MultiGML: MultiGML_Model and MultiGML_Code.