Project description
Biomolecular condensates, membraneless compartments formed by phase separation of proteins and RNA, organise biochemistry in modern cells and are plausible models for compartmentalisation in early life. The molecular rules that determine their formation, internal rchitecture and selectivity remain poorly understood. In this project, we use coarse-grained molecular dynamics at single amino acid and nucleotide resolution to link sequence to condensate behaviour. We first improve the protein model with a machine-learning (Gaussian-process/Bayesian optimisation) workflow calibrated against experimental dimensions of a large set of intrinsically disordered proteins, and benchmark it against other state-of-the-art models. We then apply the model to RNA-protein systems of increasing complexity: multiphase core-shell organisation of the nucleolus, protein relocalisation to nuclear speckles (interstasis), phosphorylation-controlled condensation versus genome compaction by the SARS-CoV-2 nucleocapsid protein, and mRNA export through the nuclear pore complex. Together, these studies aim to identify how sequence, charge patterning, stoichiometry, RNA length and post-translational modification govern RNA-protein phase separation, with implications for primitive compartmentalisation, cellular organisation and disease.


