(360bb) Benchmarking Martini 3.0 Force Field for Reproducing Thermodynamic Properties of Biomolecular Condensates

Multivalent and weak interactions between chains of intrinsically disordered proteins (IDPs) lead to formation of biomolecular condensates having liquid-like properties. It has been recently shown that during the transcription process, RNA Polymerase is recruited to promoter sites via a similar condensation mechanism and such transcriptional condensates have been found to have implications in cancer. Although atomistic molecular dynamics (MD) can provide high resolution insights into the liquid-liquid phase separation of IDPs, its high computational cost still requires coarse-graining the system to decrease the degrees of freedom. We aim to perform coarse-grained (CG) MD simulations of transcriptional condensates using the Martini force field. However, it has been shown that Martini 3.0, a CG force field, overestimates the protein-protein interactions. To address this issue, in this work, we rebalance the force-field using the low-sequence complexity domains of FUS and TDP-43 as a benchmarking system. By using deep learning, the obtained thermodynamic and transport properties from these simulations will be compared with experimental data to optimize the scaling parameters for the Martini 3.0 force field.