Current outcome: My sole-authored computational study was published in In Silico Pharmacology on 18 August 2026. Its DOI and PubMed record are linked below.
Morroniside looked like the strongest candidate when I began this study. That judgement rested on a preliminary analysis. Once I added independent runs, its ranking did not hold. Catalpol and geniposide were more consistent from one run to the next, so I dropped the original ranking and kept those two for follow-up experiments.
What I compared
The Keap1–Nrf2–ARE pathway is involved in cellular defence against oxidative stress. I compared five natural products—catalpol, geniposide, loganin, morroniside, and coptisine—against human Keap1. I was looking for non-covalent retention near the ETGE pocket that remained consistent across independent runs.
The calculations could show a retention pattern. They could not establish binding in a biochemical assay or activity in a cell. I used them only to set an order for follow-up experiments.
How the repeats changed the ranking
I applied the same docking and all-atom molecular dynamics procedure to all five compounds. To avoid ranking them from a single score, I ran independent repeats and examined the structural changes, pocket contacts, and displacement in each trajectory.
The compounds did not behave alike. Catalpol and geniposide showed the most consistent pocket-adjacent retention. Loganin varied between runs, and morroniside did not meet the repeatability criterion. Coptisine kept moving away from the pocket. I limited the endpoint-energy sensitivity analysis to the catalpol and geniposide complexes that passed that criterion.
Publication record and current limits
I kept the computational conditions consistent, ran independent repeats, and reassessed the shortlist. The trajectories from those repeat runs and the structural, contact, and displacement analyses are the internal basis for that decision.
The paper and PubMed record confirm that the research was published. They do not amount to independent experimental replication of the proposed binding mechanism. My conclusion in the paper goes only as far as testing catalpol and geniposide first. Neither compound was established as a direct Keap1 binder or an Nrf2 activator.
Where I stopped the interpretation
When the repeats failed to support the preliminary ranking, I abandoned the interpretation centred on morroniside. I also saw shared patterns in the protein's global motion, but the study had no apo control. I did not interpret those patterns as ligand-specific allostery.
The next step would be to measure direct binding and competition with the ETGE peptide. If those results support the proposal, target engagement and downstream pathway activity can then be tested in cells. Those experiments are proposals in the paper. I have not performed them.
Paper and methods
- Article DOI
- PubMed record
- First published in In Silico Pharmacology 14, article 211 (2026) by Springer Nature.
- Author: Byungwoong Yoo, Independent Researcher
- ORCID: 0009-0002-1797-3100
After molecular docking and all-atom molecular dynamics, I analysed the resulting structures and interactions. Every candidate went through the same workflow, and consistency between independent runs carried more weight than the result of any single run. The published Methods and Supplementary Information contain the detailed computational conditions.