A machine learning approach is used to analyse multi-omics (proteomics, metabolomics and transcriptomics) data, producing genetic scores for more than 17,000 biomolecular traits in human blood, and identifying possible associations with disease.