Authors
Yachen Xie, Juewen Liu
Published in
Chembiochem : a European journal of chemical biology. Volume 27. Issue 14. Pages e70481. Jul 29, 2026.
Abstract
Molecular docking is increasingly used to infer aptamer-target interactions, yet most studies rely on computationally predicted aptamer structures rather than experimentally determined ones. Using a benchmark set of aptamers with known high-resolution structures, we show that commonly used modeling approaches, including RNAComposer and AlphaFold3, fail to reliably reproduce aptamer conformations, particularly at the binding sites critical for molecular recognition. Key limitations include the use of A-form RNA models to represent B-form DNA structures, the prediction of ligand-free rather than ligand-bound conformations, and the scarcity of experimentally determined aptamer structures for training machine-learning models. Using the theophylline aptamer, for which high-resolution structures are available in both DNA and RNA forms, we systematically evaluated each step of the standard docking workflow. We found that structure-prediction errors generate incorrect binding pockets, docking scores fail to distinguish theophylline from caffeine despite a 250,000-fold difference in affinity, and molecular dynamics simulations do not overcome these shortcomings. Together, these results reveal fundamental weaknesses in current aptamer docking workflows and caution against using docking-derived models to infer binding mechanisms in the absence of experimental structural data.
PMID:
42496030
Bibliographic data and abstract were imported from PubMed on 13 Oct 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 2
- Comments 0