In Silico Modeling and Scoring of PROTAC-Mediated Ternary Complex Poses

Abstract

Proteolysis targeting chimeras (PROTACs) are bifunctional molecules that can induce the ubiquitination of targeted proteins via the formation of ternary complexes between an E3 ubiquitin ligase and a target protein. The poly-ubiquitinated target protein will be escorted to the proteasome for degradation. Rational design of PROTACs require knowledge of an accurate configuration of the PROTAC induced ternary complex. This study demonstrates that native ternary poses can be distinguished by scoring candidate poses based on the pose residence time. The scoring is essentially heat-and-dissociate trials of candidate poses sampled by MD and pre-ranked by the classic MM/GBSA method. It is practical, simple to use and self-intuitive, relying on the observation that the assumed more stable native crystal ternary poses maintained a longer residence time than non-native ones at both room and higher temperatures. A time score and temperature score were generated from multiple replicate trajectories. These scores were able to correctly identify the native pose from non-native ones in all the systems examined. The absolute numbers were comparable across different systems in all currently available VHL and CRBN-containing ternary crystal structures. Therefore, it is also possible to provide an empirical criteria for unresolved ternary structures that under the conditions of this study. If a ternary pose is stable up to over a certain threshold score, it is likely a native pose. The success of the method is in part attributed to the dynamic nature of the pose change analysis which naturally involves entropic effects, one that is intrinsically unavailable with faster static scoring methods that consider molecular mechanical energy only. Protein-protein binding entropy is much more significant than in protein-ligands binding. The success is also attributed to the fact that the protein structures themselves were all stable in the short heating trials.

PROTACable Is an Integrative Computational Pipeline of 3-D Modeling and Deep Learning To Automate the De Novo Design of PROTACs

“…The top 20 modeled poses from each ternary complex were evaluated for Cα-RMSD and L-RMSD against the crystallized complex. The lowest RMSD solutions, among the top 20 PROTAC-productive poses, for the eight popularly benchmarked ternary complexes ,− are shown in Figure .…”

“…The compared methods in Table resemble our pipeline in modeling the ternary complex starting from ternary complex building blocks, especially when it comes to resorting to protein–protein docking in order to obtain ternary complex poses. However, those methods require constrained templates of known ternary complex conformations , or molecular dynamics simulation . In contrast, our “sum-of-parts” integrative approach does not rely on any a priori knowledge of ternary complexes and starts directly from the ligand-unbound forms of the proteins.…”

“…Liao et al proposed a molecular dynamics (MD)-based strategy and heating-accelerated pose change (HOPAC) rescoring method, which requires high computational resources and manual intervention. 12 Zaidman et al provide an automated pipeline (PRosettaC), particularly at the linker insertion stage, although their method requires expert tuning based on selected atom identifiers within the PROTAC SMILES string. 14 Weng et al's approach necessitates of an established holo-form of protein complexes and user intervention for pose selection.…”

“…Recently, a range of computational methods have emerged to help the discovery of innovative PROTACs. ,− However, these methods often require substantial resources and an intricate grasp of the target’s tertiary or quaternary structure. Moreover, earlier research − predominantly validated these methods only on active crystallized PROTACs, thereby curbing their broader applicability.…”

“…Recently Liao et al, Zaidman et al, and Weng et al . leveraged computational tools to model ternary complexes.…”