Elucidating diffusion dynamics of DNA-binding proteins using physics-based simulations and Al-based modeling

Jejoong Yoo^1*

¹School of Computational Sciences, Korea Institute for Advanced Study, Seoul, Korea

* Presenter:Jejoong Yoo, email:jejoong@kias.re.kr

Intrinsically disordered regions (IDRs) are abundant in DNA-binding proteins and are increasingly recognized as crucial regulators of target search and recognition. However, standard molecular dynamics (MD) simulations fail to capture experimentally consistent diffusion dynamics, and AI-based structure prediction methods such as AlphaFold remain incapable of providing meaningful conformational ensembles for IDRs. Here, we introduce CUFIX-AMBER, a refined force field with fine-tuned nonbonded parameters for hydrogen bonding and electrostatics, which enables realistic simulations of facilitated diffusion. Using CUFIX-AMBER, we examined two representative IDR-containing proteins: the engrailed homeodomain and apurinic/apyrimidinic endonuclease 1 (APE1). Millisecond-scale simulations revealed that N-terminal IDRs not only stabilize transiently bound states but also promote dynamic alternation between sliding and hopping along DNA. Furthermore, we discuss how these state-of-the-art simulation datasets could be integrated into generative AI frameworks to move beyond static predictions and toward ensemble-level modeling of disordered protein behavior. Together, this work underscores the functional importance of IDRs in DNA search mechanisms and establishes a foundation for simulation-informed AI models.

Keywords: Molecular dynamics , DNA-binding, Diffusion, Intrinsically disordered