High-throughput sequencing of genomes of one or many species living in distinct habitats has opened extraordinary perspectives for predicting species responses to climate change. The objective of this research project is to develop statistical methods for predictive ecological genomics (PEG) based on environmental and genomic DNA. The project focusses on predicting the vulnerability of a species or an assemblage of species by using projections of altered environments. The proposed algorithms are based on modern statistical approaches, including deep latent variable models and approximate Bayesian inference. In collaboration with population geneticists and ecologists, the project will apply the methods to African crops and to Alpine biotopes threatened by global warming. This project will allow us to better understand population and ecological community responses to climate change, target plant cultivars, and assist conservation of key or endangered species.