Scoring close linker analogues in PROTAC ternary complexes: A matched-pose study of VHL-recruiting SHP2 degraders
- 1Department of Chemistry
- 2Northern University Bangladesh
Abstract
Linker optimisation in targeted protein degradation remains largely empirical, and a computational method able to rank close analogues before synthesis would be valuable. Whether current methods can do so has not been tested against the smallest chemical difference that changes potency. Here two VHL-recruiting SHP2 degraders that differ by exactly one methylene, P9 and P10, were compared across 27 ternary complex arrangements generated for both compounds, so that in every comparison the protein–protein geometry was held fixed and only the degrader varied. RxDock interaction scores did not distinguish them (p = 0.66). Prime MM-GBSA favored P9 by 8.1 kcal mol−1, agreeing in direction with the reported potency order (p = 0.040), but the preference was overturned by removing any of twelve individual poses and arose entirely from the generalized-Born solvation term; the nonpolar contribution was zero (−0.01 kcal mol−1, p = 0.97). Interface geometry computed from coordinates alone was indistinguishable between the two ensembles, with buried surface areas differing by 0.23%, and the computed energies were uncorrelated with interface size across a 2.9-fold range (r2 ≤ 0.10). Two further observations emerged: rigid-body docking score did not predict ternary competence, since none of the highest-scoring 1% of arrangements accommodated either degrader, and the longer linker reached ten more arrangements while closing no better within those both could bridge. End-point scoring of single minimized ternary complexes therefore does not resolve a one-methylene linker difference, even when it recovers the correct direction, and conformational averaging rather than a different energy function is the most likely route to improvement. The ternary complex models reported here are the first for this series.
Related
- COMPASS: A Computational Pipeline to Identify Linkers Predicting Ubiquitinable PROTAC‐Induced Ternary Complexes
- Native Mass Spectrometry Reveals Distinct Effects of Linker Length and Targeting-Ligand Identity on PROTAC Ternary Complexes
- Sub-stoichiometric degradation is dispensable for potent PROTACs as demonstrated by irreversible covalent BTK degraders
- Methods to Study the Molecular Mechanism and Drive the Design of Protein Degraders
- Cellular Morphological Profiling of PROTAC Degraders of the m6A RNA Methyltransferase METTL3.
BibTeX
@misc{islam2026scoring,
title = {{Scoring close linker analogues in PROTAC ternary complexes: A matched-pose study of VHL-recruiting SHP2 degraders}},
author = {Tasniqul Islam},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15009713/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009713/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Scoring close linker analogues in PROTAC ternary complexes: A matched-pose study of VHL-recruiting SHP2 degraders
AU - Tasniqul Islam
PY - 2026
DA - 2026/10/01
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009713/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009713/v1
AB - Linker optimisation in targeted protein degradation remains largely empirical, and a computational method able to rank close analogues before synthesis would be valuable. Whether current methods can do so has not been tested against the smallest chemical difference that changes potency. Here two VHL-recruiting SHP2 degraders that differ by exactly one methylene, P9 and P10, were compared across 27 ternary complex arrangements generated for both compounds, so that in every comparison the protein–protein geometry was held fixed and only the degrader varied. RxDock interaction scores did not distinguish them (p = 0.66). Prime MM-GBSA favored P9 by 8.1 kcal mol−1, agreeing in direction with the reported potency order (p = 0.040), but the preference was overturned by removing any of twelve individual poses and arose entirely from the generalized-Born solvation term; the nonpolar contribution was zero (−0.01 kcal mol−1, p = 0.97). Interface geometry computed from coordinates alone was indistinguishable between the two ensembles, with buried surface areas differing by 0.23%, and the computed energies were uncorrelated with interface size across a 2.9-fold range (r2 ≤ 0.10). Two further observations emerged: rigid-body docking score did not predict ternary competence, since none of the highest-scoring 1% of arrangements accommodated either degrader, and the longer linker reached ten more arrangements while closing no better within those both could bridge. End-point scoring of single minimized ternary complexes therefore does not resolve a one-methylene linker difference, even when it recovers the correct direction, and conformational averaging rather than a different energy function is the most likely route to improvement. The ternary complex models reported here are the first for this series.
ER -