Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.64898/2026.07.10.737860

Benchmarking AI Protein Structure Predictors Reveals a Persistent Bias in Multi-State Proteins

Ye, M.·Wang, Y.-H.·Brogi, M.·Parks, J. M.·Kuo, K. M.·Gumbart, J. C.
View original source ↗Export metadata ↓

Abstract

Protein structure predictors achieve high single-state accuracy, but it remains unclear whether they can recover functionally relevant conformational ensembles or account for the presence of ligands and/or binding partners. Here, we benchmark AlphaFold3, Boltz-2, Chai-1, and BioEmu on four canonical multi-state proteins (Pf-MATE, LAO, SecA, and {beta}2AR), quantifying state bias and sampling breadth against experimental reference structures. Models frequently default to a dominant state represented in the PDB; small-molecule ligands have weak or inconsistent effects, while large protein partners drive clear conformational switching between states. Multiple sequence alignment (MSA)-based approaches (AF-Cluster and random subsampling) recapitulate similar biases, indicating that this behavior is not unique to newer architectures. These results underscore current limitations for multi-state protein structure prediction and structure-guided ligand discovery. TOC Graphic O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=111 SRC="FIGDIR/small/737860v1_ufig1.gif" ALT="Figure 1"> View larger version (12K): org.highwire.dtl.DTLVardef@278cc1org.highwire.dtl.DTLVardef@8a0b66org.highwire.dtl.DTLVardef@f28a21org.highwire.dtl.DTLVardef@14ab675_HPS_FORMAT_FIGEXP M_FIG C_FIG

Explore related subjects

biophysicsPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Ye, M., Wang, Y.-H., Brogi, M., Parks, J. M., Kuo, K. M., Gumbart, J. C.. 2026-07-11. Benchmarking AI Protein Structure Predictors Reveals a Persistent Bias in Multi-State Proteins. https://doi.org/10.64898/2026.07.10.737860

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

A quantum state of mitochondria in the living cell

The high energy-efficiency of life is hard to understand only with classical physics. Many efforts have been made to study its mechanism based on quantum mechanics; the progress is nevertheless slow due to lack of experimental evidence with living cells. Here, combining experiments on cells, tissues and mitochondria with a theoretical model, we demonstrate a quantum state of mitochondria, which can be employed to modulate ATP production in living cells. We found an anomalous 71.0-THz oscillation mode only in living cells and tissues, which is highly determined by intact structure of mitochondria, and cannot be assigned to any specific molecules. Based on experimental data, a quantum model of light-matter coupling was introduced to trace the origin of this mode. Our calculations suggest a quantum superposition state of functional mitochondrion that forms by the coupling of light and lipid CH2 bonds in functional cristae, and induces a splitting of the intrinsic CH2 vibration mode of 87 THz to two levels at 71 THz and 103 THz, respectively. The former can be observed only in living cells and tissues; whereas the latter falls in the range (90-110 THz) of biomolecular and water vibrations, thus indistinguishable. Additional experiments revealed this mitochondrial quantum state able to serve as an efficient channel to modulate ATP production. Our findings provide a quantum mechanics view for understanding living cells, and it will be interesting to further explore whether such quantum state could act as a channel for energy metabolism, and even information transmission in life.

biophysics↗
bioRxiv2026

Mechanistical and structural basis of Kv channel inhibition by 4 aminopyridine

Inhibition of Kv channels by 4-aminopyridine (4AP) improves motor function in multiple sclerosis by enhancing neuronal excitability. The mechanism of inhibition and the structural basis of 4AP binding to Kv channels remain unclear. Here, we determined the structure of the Shaker V369I-I372L-S376T (ILT) mutant bound to 4AP at 3.3 [A], demonstrating that 4AP binds to the closed state of the channel. This structure is inconsistent with an open channel block mechanism. Electrophysiology experiments show that 4AP binds even when intracellular pore access is constitutively blocked, suggesting that 4AP enters the pore through membrane-facing fenestrations. MD simulations and mutational analysis agree with the proposed fenestration pathway and suggest that 4AP binds in its neutral form. These results support a mechanism where 4AP binds to a partially activated closed state that prevents complete activation of Kv channels, explaining its pharmacological activity.

biophysics↗
bioRxiv2026

Substrate binding reorganizes the energetic landscape of Plasmodium falciparum hexose transporter PfHT1

Malaria parasites depend on the Plasmodium falciparum hexose transporter PfHT1 for sugar uptake, yet how substrate binding reshapes transporter energetics and kinetics of sugar transport remains poorly understood. Here, we investigate how glucose reorganizes the conformational landscape, transition pathways, and residue interaction network underlying membrane transport. Using over 800 s of adaptive molecular dynamics simulations combined with Markov state models, transition-path theory, residue-contact analysis, and graph attention learning, we reconstruct the apo and glucose-bound conformational cycles. We show that glucose selectively stabilizes productive outward-facing, occluded, and inward-facing conformations, reshapes transition kinetics, and channels reactive flux through the occluded state. We identify TM7b helix cracking as a local structural transition coupled to extracellular-gate closure and substrate progression, providing a flexible connection between the binding pocket and global alternating access. Experimental testing of mechanistically critical residues validated their functional importance in PfHT1-dependent sugar utilization. Together, these results show how substrate binding reorganizes the energetic, kinetic, and interaction landscape of membrane transport and establish a transferable framework for studying transporter mechanisms.

biophysics↗