PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 15, 20260 citationsOpen Access

Lo-Shu/Markov-Fiedler Framework v4: Full Size-Controlled Allosteric Benchmark (n=569), Multi-Target Vina Discrepancy (n=46), and Honest Positioning

View Full Paper
YKYao-Kai Kao

Key Points

  • To validate the Lo-Shu/Markov-Fiedler C-alpha geometric framework for allosteric classification and multi-target binding.
  • Analyzed n=569 proteins for allosteric classification, correcting size confound using OLS residualization.
  • Evaluated multi-target Vina discrepancies across 46 ligands and 8 targets.
  • Conducted performance analysis of models using Random Forest AUC.
  • After correction, significant factors included lambda_2_res (p=0.04), flex_res (p=0.003), and rig_res (p=0.001).
  • Random Forest AUC achieved 0.882 with a permutation p-value of 0.005.
  • 38 of 46 ligands scored under target expectations, supporting the domain-boundary hypothesis for kinases.

Abstract

Full validation of the Lo-Shu/Markov-Fiedler C-alpha geometric framework across three tasks. (1) Allosteric classification on n=569 proteins (486 allosteric from UniProt KW-0021+ASBench, 83 orthosteric-only). Size confound (rho (CHC, n) =0. 860) confirmed and corrected via OLS residualization. After correction: CHC loses significance; lambda₂ᵣes (p=0. 04), flexᵣes (p=0. 003), rigᵣes (p=0. 001) retained. RF AUC=0. 882 (permutation p=0. 005). (2) Multi-target Vina discrepancy: 8 targets, 46 ligands, 37/46 under-scored; domain-boundary hypothesis supported for kinases, not HIV-PR (honest negative). (3) Calibrated claim: original lightweight integration, not world-first. Next step: run PASSer/ALLO on same dataset for direct head-to-head comparison.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Yao-Kai Kao (2026) studied this question.

synapsesocial.com/papers/6a06b914e7dec685947aba48https://doi.org/10.5281/zenodo.20152192
Ask AI
Helpful
Bookmark
Share
View Full Paper