Quantitative Methods and Machine Learning Models in Evolutionary Plant Population Genomics for Analysis of Variant Effects

The PhD project uses biological language models, deep learning and quantitative genomics methods to propose novel methods for predicting variant effects at single-base resolution in plant genomes and evaluates them in temperate grass species. Work Package 1 will introduce a supervised approach to predict nucleotide conservation in protein-coding regions, based on multi-genome alignments at different evolutionary timescales. Work Package 2 will evaluate existing and novel methods to predict nucleotide conservation in non-coding regions, specifically by incorporating information about gene regulation. Work Package 3 will evaluate the potential of biological language models to infer gene co-expression and predict the effect of DNA or protein variants on gene co-expression.