Title:	Data-driven modeling of protein phase separation at cell membranes
DNr:	NAISS 2026/3-355
Project Type:	NAISS Medium
Principal Investigator:	Giulio Tesei <giulio.tesei@mau.se>
Affiliation:	Malmö universitet
Duration:	2026-04-28 – 2027-05-01
Classification:	10407
Homepage:	https://mau.se/en/persons/giulio.tesei/
Keywords:

Abstract

Biomolecular condensates (BMCs) are dynamic assemblies of primarily proteins and nucleic acids that compartmentalize the cellular environment. Recent experiments indicate that BMCs may also be enriched in phospholipids and aid their phosphorylation in signaling pathways associated with cell proliferation and apoptosis. Moreover, receptors and other membrane-associated proteins, along with their binding partners, have been shown to form BMCs, potentially enhancing signal transduction by increasing local concentration and dwell times of signaling proteins. Hence, BMCs may act as microdomains that promote phospholipid metabolism and signaling both in the cytosol and at membrane interfaces. Disruptions in the formation or dissolution of such BMCs have been linked to conditions such as insulin resistance and Alzheimer’s disease. Coarse-grained (CG) molecular models with residue-level resolution, such as CALVADOS, provide quantitative predictions of BMC formation from protein sequence and have been instrumental in interpreting experimental observations on BMCs. For single-component systems of intrinsically disordered regions (IDRs), which lack a well-defined folded structure, CALVADOS simulations have also enabled the training of machine-learning models to predict conformational and phase properties at the proteome scale, revealing links between amino acid sequence, phase separation, and protein function. BMCs involved in signaling often assemble at or near membranes and interact with lipids, but current residue-level CG models focus primarily on simplified systems that do not account for the complexity of cell membranes and the chemical diversity of phospholipids. As a result, computational tools capable of modeling these multicomponent, multiphase systems remain limited. Building on recent advances, this project aims to close this gap by developing data-driven models that capture the interplay between proteins and lipids in membrane-associated BMCs. The project has the following objectives: 1. Leverage experimental data to develop and validate a CG model for membrane proteins and cell membranes. 2. Perform simulations across long length and time scales to explore the effects of interactions between BMCs and cell membranes on their stability and material properties. 3. Develop machine-learning models trained on simulation data to predict phase behavior of membrane associated proteins as a function of amino acid sequence and lipid composition.