In the human genome, rare variants can play an important role in disease risk. Understanding these variants may help to discover individual predisposition to complex diseases and their heritability. Our aim is to develop interpretable generative modelling strategies that allow to capture the effect of rare genetic variants on gene expression. Technically, this project will entail latent variable modelling of single and multiple omics modalities as well as embeddings of genetic risk factors. We aim to identify linkages between rare genetic effects and molecular states across different cells.