Biomolecular condensates (condensates) are complex subcellular structures that emerge from the liquid-liquid phase separation of intrinsically disordered proteins (IDPs). Condensates have been implicated across both normal cellular function and disease. Given that the processes by which biomolecular condensates form are not fully understood, it is of interest to elucidate how the sequence composition of constituent IDPs influence their material properties and underlying physical behavior. However, modeling condensate systems via standard molecular dynamics (MD) techniques is challenging, as simulating under the canonical ensemble induces finite-size effects (FSE) which alter system thermodynamics and make accurately reproducing in vivo conditions difficult. In this thesis, I propose a methodology for predicting relevant kinetic and thermodynamic parameters of condensates in the macroscopic limit by applying the modified liquid droplet (MLD) framework to ensembles of finite-size MD simulation trajectories. To this end, I simulated three variants of heterogeneous nuclear ribonucleoprotein A1 low-complexity domain (HNRNPA1-LCD) across multiple system densities and volumes to calculate FSE-resolved nucleation barrier height, critical nucleus size, surface tension, and dilute phase density. I find that the emergent properties of HNRNPA1-LCD condensates are modulated by sequence variation, and that their observed behavior can be rationalized from the physicochemical properties of the introduced mutations. In addition, I find that the performance of the MLD framework is sensitive to simulation conditions, as high-density, low-volume regimes yield especially severe FSE that are difficult to resolve.

Biomolecular condensates are stabilized by the collective action of short, low-affinity interactions distributed across intrinsically disordered protein sequences, precisely the interaction grammar that short peptides are chemically designed to engage. This conceptual alignment makes tetrapeptides and other minimal sequence fragments natural candidates for condensate modulation: they are readily synthesized, chemically tunable across a vast combinatorial space, and small enough to partition into the dense phase without the steric constraints that limit larger molecules. Yet whether short peptides act as passive occupants of the condensate milieu or as active, sequence-specific perturbants of its internal architecture remains an open question. Here, we use Mpipi coarse-grained molecular dynamics simulations to characterize how a combinatorial library of tetrapeptides drawn from a low-dimensional sequence space actively perturbs two prototypical LCD condensates: hnRNPA1-LCD and FUS-LCD. We find that tetrapeptide partitioning into the condensate is governed primarily by aromatic hydrophobicity, which drives exponential increases in uptake, while charge plays a more delicate secondary role. Also, the spatial mode of perturbation is controlled not by peptide composition alone but by the host protein’s aromatic sequence architecture: the same tetrapeptides produce discrete, hotspot-concentrated disruptions in A1-LCD where charged and aromatic residues are clustered, and diffuse, sequence-wide perturbations in FUS-LCD where aromatics are regularly spaced. This architectural dependence is confirmed by binding–disruption alignment analysis, where spatial coupling between peptide binding and contact remodeling is strong in A1-LCD and collapses in FUS-LCD. Despite this local remodeling, the condensates’ hub–clique network topology remains invariant across all conditions, indicating it is an emergent collective property of the condensed state rather than a tunable structural feature.