Small molecules mostly shifted between overlapping sets of protein residues rather than making long jumps, according to a computational study that represents binding as a chain of stochastic contact states. In the main Aβ42 analysis, long-range hops accounted for approximately 7% of transitions, while most transitions involved residue sets with a Jaccard similarity of at least 0.66, a score that captures how much two sets overlap.
The study asks whether a contact-state Markov state model can represent small-molecule binding to disordered proteins as stochastic binding paths and quantify affinity. It follows changes in residue contacts as transitions between states, giving the analysis a way to describe a route into binding rather than only an endpoint.
It examined two systems: 10074-G5 with Aβ42 and fasudil with α-synuclein. For the main Aβ42 analysis, VAMPNet used ligand-Aβ42 distance time series from a 28 μs molecular-dynamics simulation.
A trail of overlapping contacts
The hierarchical VAMPNet attempt to divide the Aβ42 trajectory into discrete states exposed a problem. It produced 16 states from four clusters at each of two levels, but more than 90% of frames mapped to state 2. No comparable kinetically separable bound configurations were found, so conventional clustering did not yield a balanced set of binding states.
The analysis then moved to residue-level handoffs. Its Aβ42 handoff matrix was 42 by 42, with each entry counting a transition from a ligand contact at one residue to a contact at another residue at the next time step. That set up the model's distinction between gradual, overlapping moves and longer-range hops.
The gliding pattern was not limited to the Aβ42 trajectory. Across 45 IDP-ligand trajectories spanning 12 diverse intrinsically disordered proteins, high-overlap gliding transitions accounted for most of the modeled movement between states in every system.
Testing the model
To test the model's consistency across time steps, the study used Chapman-Kolmogorov validation. The quintuplet MSM contained 13,034 states and showed the best agreement among the tested cluster sizes.
That large state network was then reduced by Louvain clustering from 13,034 states to 146 macrostates.
The validation was not perfect. Systematic deviations persisted at longer time steps, and confidence bands for smaller n-plets were wide; the analysis attributed those bands to limited statistical power. The MSM methods also assume Markovian, time-independent, linear dynamics with constant transition rates.
Affinity estimates, with limits
Those caveats matter when the model is used to estimate affinity. For Aβ42-G5, frame occupancy gave a dissociation constant, K_D, of 49 ± 4 μM. The ±4 μM uncertainty came from 10 repeated sliding-window calculations. The MSM estimate was 47 μM, while the study quoted experimental values of 6 and 40 μM.
For α-synuclein, the estimates varied by system and ligand. In three C-terminal systems, equilibrium K_D values were 25, 26 and 18 μM for ligands 40, 50 and 12, respectively. Full-length α-synuclein with fasudil yielded 60 μM.
The model also pointed to a mixed structural picture for the higher-affinity states. They showed elevated hydrophobicity, reduced solvent-accessible surface area for the contact set and modest transient local ordering. The reported Aβ42 regions were residues 17 to 21 and 30 to 41, but no single feature dominated.
The authors frame binding paths, kinetic matching and path features as possible design considerations for selectivity. That is a proposal about how the framework might be used, rather than an outcome demonstrated by the affinity calculations.
Taken together, the results describe a computational map built from molecular-dynamics trajectories and model-defined contact states. The broader test across 45 trajectories and 12 disordered proteins supports the recurrence of the gliding pattern in the systems examined, while the model's assumptions and the remaining Chapman-Kolmogorov deviations leave its wider reach unresolved.
Paper data and sources
Original title: Binding without Docking: Small-Molecule Binding Paths on Disordered Proteins
Authors: Louet A, Hummer G, Vendruscolo M
Journal/Repository: Journal of the American Chemical Society
Status: Peer-reviewed
First online: 2026-10-19
DOI: Not available
Original paper · Full text