Related Experiment Video
Updated: Oct 15, 2026

A Bilingual Computational Workflow for Identifying Potential PLK1 Inhibitors in American Sign Language and English
Published on: April 3, 2026
Knowledge graph and machine learning-guided virtual screening prioritizes putative blockers of the PSMD11/Rpn6-EBOV
Yanshuang Wang1, Yuanjian Li2, Lin Zhu1
1National Key Laboratory of Advanced Biotechnology, Academy of Military Medical Sciences, Beijing, China.
Introduction:
Host-directed antiviral discovery requires a traceable route from host-factor identification to testable molecular hypotheses. This study developed such a workflow around a host factor associated with Ebola virus glycoprotein (EBOV GP).
Methods:
GFP-GP pulldown coupled with co-immunoprecipitation/mass spectrometry was used to identify GP-associated host candidates. A PSMD11/Rpn6-centred workflow integrated knowledge-graph inference, ligand-only machine learning, ADMETlab profiling and interface-guided docking. Candidate prioritization was assessed through score-weight sensitivity analysis, applicability-domain assessment and docking controls addressing seed reproducibility, search convergence, whole-receptor context and macrocycle ring-conformer sensitivity. Docking scores were contextualized against a matched 59-compound library.
Results:
The analysis recovered 424 GP-pulldown-specific host proteins and nominated PSMD11/Rpn6 as a proteostasis anchor. The tier-gated knowledge-graph candidate membership was invariant across the tested weight perturbations, with Spearman rho = 0.998 under equal weighting. The selected ExtraTrees model achieved an AUC of 0.912 for the Ki endpoint. Applicability-domain assessment and ADMET profiling differentiated candidate confidence and predicted liabilities. Docking reproducibility checks yielded a maximum standard deviation of 0.05 kcal/mol for conformationally restricted ligands, while macrocyclic candidates showed greater protocol-dependent variability. Rule-based triage integrating upstream evidence, ADMET burden, chemotype diversity and mechanistic coverage returned five candidates: betamethasone, bafilomycin A1, everolimus, pitavastatin and CHEMBL1410015.
Discussion:
The workflow links a proteomic host-factor signal to a compact set of chemically diverse hypotheses with explicit evidence sources and limitations. The candidates remain hypotheses for experimental assessment; direct target engagement, interface blockade and antiviral activity are not established by the present study.
Related Concept Videos
Protein-protein Interfaces
Pharmacogenomics: Identification of New Drug Targets
Inhibitors of Viral Protein Synthesis
Drug Discovery: Overview
