Fig. 1: Main objectives.

Throughout the study, we generate a Disease Molecular Similarity Network containing directed positive and negative Relative Molecular Similarity (pRMS and nRMS) interactions (a). pRMS are validated using epidemiological networks extracted from medical and hospital claims (b). We stratify patients into subgroups and calculate subgroup-specific relative molecular similarities (c), identifying similarities that were not recovered when working at the disease level (d), including some interactions opposing general trends (e). We propose biological processes potentially involved in subgroup-similarities (f). Finally, for each single patient, we provide a ranked list of most similar to less similar diseases based on transcriptomics information (g).