Deep generative design and structure-based evaluation of celecoxib-inspired PTGS2/COX-2 ligands: a computational study in the context of adenomyosis-associated inflammation

Adenomyosis is a chronic inflammatory gynecological disorder in which cyclooxygenase-2 (PTGS2/COX-2)-derived prostaglandin signaling contributes to inflammation. This study asked whether deep learning-based generative modification of the celecoxib scaffold yields compounds with a more favorable predicted interaction profile at PTGS2. A generative platform combining a three-dimensional convolutional neural network with a conditional variational autoencoder returned twelve analogues, all carried forward without post hoc selection and examined by substructure analysis, docking at the experimentally defined ligand-binding site, pharmacophore comparison, ADMET and toxicity prediction, density functional theory, and 200 ns molecular dynamics with MM/GBSA estimation. The benzenesulfonamide group of celecoxib was retained in five of twelve analogues and the trifluoromethyl-pyrazole core in none, so the compounds are better described as scaffold-hopped analogues than as close derivatives. AI Analogue 2 returned the most favorable predicted binding energy (− 10.024 kcal/mol; celecoxib − 7.748 kcal/mol) and formed a persistent complex with an effective binding energy of − 51.89 kcal/mol. Its predicted pose lies along the main cyclooxygenase channel rather than the COX-2 side pocket occupied by celecoxib, so the two energies are not comparable as affinities. Redocking of the co-crystallized ligand gave a deviation of 1.924 Å and the celecoxib trajectory did not converge, indicating a rigid-receptor starting-pose artifact rather than a property of the generated compounds. Predicted blood–brain barrier permeability coincided with an active neurotoxicity call, identifying central exposure as the principal predicted liability. The study does not establish COX-2 inhibition, selectivity over COX-1, improved safety, or therapeutic activity in adenomyosis .

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-16
DOI
https://doi.org/10.1038/s41598-026-75822-4
Primary Topic
Computational Drug Discovery Methods
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Deep generative design and structure-based evaluation of celecoxib-inspired PTGS2/COX-2 ligands: a computational study in the context of adenomyosis-associated inflammation

Nouman Safdar Ali, Sibel Üstünel
Scientific Reports
Computational Drug Discovery Methods
article

Deep generative design and structure-based evaluation of celecoxib-inspired PTGS2/COX-2 ligands: a computational study in the context of adenomyosis-associated inflammation

Nouman Safdar Ali, Sibel Üstünel
article en

Abstract

Adenomyosis is a chronic inflammatory gynecological disorder in which cyclooxygenase-2 (PTGS2/COX-2)-derived prostaglandin signaling contributes to inflammation. This study asked whether deep learning-based generative modification of the celecoxib scaffold yields compounds with a more favorable predicted interaction profile at PTGS2. A generative platform combining a three-dimensional convolutional neural network with a conditional variational autoencoder returned twelve analogues, all carried forward without post hoc selection and examined by substructure analysis, docking at the experimentally defined ligand-binding site, pharmacophore comparison, ADMET and toxicity prediction, density functional theory, and 200 ns molecular dynamics with MM/GBSA estimation. The benzenesulfonamide group of celecoxib was retained in five of twelve analogues and the trifluoromethyl-pyrazole core in none, so the compounds are better described as scaffold-hopped analogues than as close derivatives. AI Analogue 2 returned the most favorable predicted binding energy (− 10.024 kcal/mol; celecoxib − 7.748 kcal/mol) and formed a persistent complex with an effective binding energy of − 51.89 kcal/mol. Its predicted pose lies along the main cyclooxygenase channel rather than the COX-2 side pocket occupied by celecoxib, so the two energies are not comparable as affinities. Redocking of the co-crystallized ligand gave a deviation of 1.924 Å and the celecoxib trajectory did not converge, indicating a rigid-receptor starting-pose artifact rather than a property of the generated compounds. Predicted blood–brain barrier permeability coincided with an active neurotoxicity call, identifying central exposure as the principal predicted liability. The study does not establish COX-2 inhibition, selectivity over COX-1, improved safety, or therapeutic activity in adenomyosis .

Scientific ReportsVol. 16(1)
Atlas Üniversitesi (TR)
Good health and well-being
Computational Drug Discovery Methods
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon