Computational and Structural Biotechnology Journal (Dec 2024)
Cyclodextrins: Establishing building blocks for AI-driven drug design by determining affinity constants in silico
Abstract
Cyclodextrins (CDs) are cyclic carbohydrate polymers that hold significant promise for drug delivery and industrial applications. Their effectiveness depends on their ability to encapsulate target molecules with strong affinity and specificity, but quantifying affinities in these systems accurately is challenging for a variety of reasons. Computational methods represent an exceptional complement to in vitro assays because they can be employed for existing and hypothetical molecules, providing high resolution structures in addition to a mechanistic, dynamic, kinetic, and thermodynamic characterization. Here, we employ potential of mean force (PMF) calculations obtained from guided metadynamics simulations to characterize the 1:1 inclusion complexes between four different modified βCDs, with different type, number, and location of substitutions, and two sterol molecules (cholesterol and 7-ketocholesterol). Our methods, validated for reproducibility through four independent repeated simulations per system and different post processing techniques, offer new insights into the formation and stability of CD-sterol inclusion complexes. A systematic distinct orientation preference where the sterol tail projects from the CD's larger face and significant impacts of CD substitutions on binding are observed. Notably, sampling only the CD cavity's wide face during simulations yielded comparable binding energies to full-cavity sampling, but in less time and with reduced statistical uncertainty, suggesting a more efficient approach. Bridging computational methods with complex molecular interactions, our research enables predictive CD designs for diverse applications. Moreover, the high reproducibility, sensitivity, and cost-effectiveness of the studied methods pave the way for extensive studies of massive CD-ligand combinations, enabling AI algorithm training and automated molecular design.
