Health Data Modelling

We all differ by few or more letter exchanges in our genetic code that makes us individual. Such naturally occurring genetic variation, called single nucleotide polymorphisms (SNPs), sometimes also makes us more likely to develop a disease or to react to medications differently. There are billions of those SNPs in the human genome but only a tiny fraction directly changes the sequence of genes that encode proteins, the work horses of our bodies. Our group uses these large-scale genetic experiments by nature for two main purposes: 

1. to identify novel targets for drugs or to repurpose existing, safe, and effective drugs for diseases without good treatments, and 

2. to identify genetic variants that may alter how drugs work in our body, possibly loosing efficacy or exerting adverse events. 

We decipher genetic risk information observed in millions of participants by integrating deep molecular data down to single cell level by utilizing machine learning. Once we identify responsible genes, and hence proteins, that account for disease risk, we can ask whether there are already existing drugs to target those proteins. For example, we provided evidence for several existing drugs to be of potential use for the treatment of the common disease Raynaud’s phenomenon, for which no good medications exist (Hartmann et al. 2023 Nature Communications). Along the same lines, profiling the molecular consequences of genetic variation can point to process important for the efficacy but also absorption, distribution, metabolism, and excretion of drugs. For example, we identified a common genetic variation affecting the function of the gene DYPD that is important for timely degradation of the cancer medication 5-fluorouracil that may otherwise reach toxic levels harming patients (Surendran et al. 2022 Nature Medicine).