Multi-representation particle-field simulations of intrinsically disordered proteins

Ritvind Suketana

doi:10.17918/00011149

Intrinsically disordered proteins (IDPs) are a widespread and functionally critical class of proteins that lack stable tertiary structure but are central to the formation and regulation of biomolecular condensates. Modeling the phase behavior and dynamics of IDPs is inherently challenging due to their sequence dependent interactions, heterogeneous conformations and multi-scale phase behavior. Traditional simulation techniques such as molecular dynamics (MD) and field theoretic simulations (FTS) offer complementary strengths but suffer from individual limitations in scale, resolution or efficiency that make it difficult for either of these methods alone to effectively simulate IDPs. This thesis introduces and validates a multi-representation particle-field framework that couples coarse grained MD to FTS, enabling rapid equilibration of mesoscale IDP condensates without sacrificing residue level resolution where dynamics matter. We first establish formal equivalence between particle and field based models in systems with explicit solvent, validating agreement across thermodynamic observables and phase diagrams -- with FTS reaching equilibrium orders of magnitude faster than MD. Leveraging this speed, equilibrated FTS density fields are back-mapped to particle coordinates to seed MD, greatly reducing the cost of dynamical studies. We apply this hybrid methodology to a simplified model of a stereotypical IDP, reproducing single-chain radii of gyration and experimental phase separation binodals with a single, self-consistent interaction matrix. However, numerical instability in FTS prevents us from extending this model to systems comprising of more than three species. To improve the stability and speed of field-based simulations and enable their application to chemically complex systems, we introduce a Bayesian optimization approach for tuning field coefficients and stabilizing high-dimensional field updates in FTS. This tool achieves up to 190-fold improvements in simulation efficiency and enables the use of FTS for multi-component, explicit solvent IDP systems that were previously intractable. Finally, sequence dependent dynamics are explored using our multi-representation framework. By systematically varying charge patterning in polyampholytic IDPs, we reveal that blocky charge sequences exhibit an inversion in single chain conformational behavior between dilute and condensed phases -- they are compact in dilute solution but become more expanded within condensed phases, a reversal driven by electrostatic screening, crowding effects and solvent behavior. While it has recently become common in literature to infer condensate properties from single chain observables, our findings highlight the need for caution when doing so, as environmental context can substantially alter molecular behavior. Overall, this work delivers a quantitatively validated, computationally efficient multi-representation toolkit for studying IDP phase behavior. It lays the groundwork for scalable, sequence-specific modeling of biomolecular condensates by integrating polymer physics, molecular simulation, and machine learning based optimization.

Multi-representation particle-field simulations of intrinsically disordered proteins

Files and links (1)

Abstract

Metrics

Details

Multi-representation particle-field simulations of intrinsically disordered proteins

Files and links (1)

Abstract

Metrics

Details

Drexel University Social media