What question did this study set out to answer?

The study aims to enhance druggability prediction and allosteric site discovery using a Markov-Fiedler framework.

May 15, 2026Open Access

Luoshu-Markov Geometric Framework: Fiedler Eigenvalue for Druggability Prediction, Allosteric Site Discovery, and Vina Discrepancy Correction (n=123)

Key Points

The study aims to enhance druggability prediction and allosteric site discovery using a Markov-Fiedler framework.
Analyzed 123 DUD-E proteins (87 druggable, 36 undruggable) using the Fiedler eigenvalue of the residue-contact graph Laplacian.
Conducted a combined model evaluation with 5-fold cross-validation to calculate AUC.
Performed discrepancy analysis for CDK2 using Vina to explain binding gap in specific interactions.
The Fiedler eigenvalue significantly distinguishes druggable (0.057) from undruggable proteins (0.090, Mann-Whitney p=0.0028).
The combined model achieved a 5-fold CV AUC of 0.8393.
CDK2 Vina analysis identified an entropic gap of +3.13 kcal in binding assessments.

Abstract

We introduce the Markov-Fiedler spectral complement to the Lo-Shu druggability framework. For n=123 DUD-E proteins (87 druggable, 36 undruggable), the Fiedler eigenvalue lambda2 of the residue-contact graph Laplacian significantly distinguishes druggable from undruggable proteins (drug=0.057 vs undrug=0.090, Mann-Whitney p=0.0028). Combined model achieves 5-fold CV AUC=0.8393. Lo-Shu Markov allosteric score recovers validated allosteric proteins (cMyc-Max, CaM, HIF-1alpha, XIAP, PTPB) from structural geometry alone. CDK2 Vina discrepancy analysis explains BMS387032 gap (+3.13 kcal) via entropic N/C lobe boundary binding.

Demander à l'IA

Bookmark

View Full Paper