Pith. sign in

REVIEW 4 major objections 5 minor 40 references

Neural spectroscopy of AlphaFold2 reveals encoded protein conformational landscapes

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read AlphaFold2's trained weights carry a readable record of protein conformational organization — stability orderings, flexibility patterns, funnel topology — that emerges under perturbation despite never being a training target.

desk verdict A serious, candid interpretability study with real external validations, but its central 'encoded landscape' claim rests entirely on a reduced-MSA regime and is not yet established. read the letter →

arxiv 2607.16087 v1 pith:JZFI4PLT submitted 2026-07-17 cs.LG q-bio.BM

Kaustav Mehta This is my paper
classification cs.LGq-bio.BM
keywords neuralspectroscopyAlphaFold2ScaledGaussianConvolutionweightperturbationconformationallandscapesproteindynamicsEvoformerinterpretability
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that AlphaFold2's 93 million trained parameters are not just machinery for turning sequence into structure: they encode a map of protein conformational organization that ordinary inference never expresses. The author's probe — scaled Gaussian convolution, which smooths each Evoformer weight tensor with a narrow Gaussian, scales it down, and applies this cumulatively to the first d of the 48 blocks — produces graded structural responses that track known physics. Ubiquitin's native contacts break in the experimentally established stability order (intermediate-boundary contacts first in all informative conditions, folding-nucleus contacts last), per-residue perturbation sensitivity correlates with microsecond-scale molecular dynamics flexibility beyond what atomic burial predicts, and the 912,384-structure ensemble reproduces the funnel topology of millisecond folding simulations. Matched-power random noise destroys the signal, and five independently trained models converge on the same landscape, arguing the organization is a property of the training, not of one weight realization. If the claim holds, conformational information is present, at least partially, inside a predictor trained only on static structures — a route to dynamics-like information without simulation.

What carries the argument

Scaled Gaussian Convolution (SGC): each weight tensor W in Evoformer blocks 0 through d−1 is replaced by W′ = λ·G_σ(W), where G_σ is Gaussian smoothing with width σ and λ is a uniform scaling factor — at the primary operating point σ = 0.30, λ = 0.75 — and d (perturbation depth) controls how many of the 48 sequential blocks are modified. Scaling dominates (more than 99% of the perturbation power) and acts as a near-uniform attenuation of every dot-product in the network, shifting information routing smoothly; the sub-1% smoothing residual has a component orthogonal to uniform scaling that selectively destabilizes specific contact groups near the transition boundary. The matched noise control

What would settle it

Repeat the SGC sweep at the default MSA depth (512×5,120) with scaling factors well below the tested λ range: if no perturbation strength reproduces ubiquitin's contact-breaking order (G2 first, G1 last), the RMSF correlations, and the funnel topology, then the physical signal is an artifact of the weakened-input regime rather than an encoding in the weights. The paper's own proposed proteome-scale screen offers the complementary check: structured landscapes should track training-data density if the account is memorization, or extend beyond explicit coverage if it is generalized conformational

Watch

Extended reading notes

Core claim

The central claim is that AlphaFold2's Evoformer weights encode physically structured conformational organization as a byproduct of the structure-prediction objective, and that this encoding can be read directly by deterministic weight deformation. Under SGC at the primary operating point, ubiquitin's native contacts lose coherence in the order established by folding experiments: in all 84 informative perturbation conditions the intermediate-boundary contacts break first, and in 67 of 84 the folding-nucleus contacts outlast every other group (ties in the remaining 17). Per-residue perturbation sensitivity tracks a 154.6-microsecond equilibrium simulation with partial correlation 0.66–0.79 af

Load-bearing premise

All primary results come from a deliberately weakened regime — MSA depth 64×64, one-eighth of AlphaFold2's default — chosen because it makes weight perturbation measurable, and the load-bearing assumption is that the physical ordering, flexibility correlations, and funnel topology arise from what the weights encode rather than from the under-supplied coevolutionary input; behavior at default depth is untested.

Editorial extensions

If this is right

  • If the readout is genuine, a static-structure predictor's weights encode a differential robustness hierarchy: contacts that are evolutionarily and physically most load-bearing are encoded most robustly, so the order in which perturbation dismantles native contacts is a readout of stability ordering.
  • The shallow-depth flexibility signal — partial correlation 0.66–0.79 beyond burial, exceeding a geometry-only network baseline — implies the weights carry a residue-level constraint signal correlated with physical flexibility, even though training never rewarded it.
  • The three-protein spectrum gives a general assay: strong training signal produces convergent landscapes, absent signal produces convergent absence, ambiguous signal produces structured disagreement — a way to determine, for any protein, where a network's representation is determined by data and where it is underdetermined.
  • The noise-control contrast establishes that the response is not generic network damage: only weight-coherent perturbation reproduces the physical signal, so the conformational information sits in the dominant, structured directions of the learned weights rather than in fine-grained random components.
  • Model-independence for ubiquitin (transition onset within two blocks across five models) implies the landscape is a property of the architecture plus training data, not of a single weight configuration.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • My extension: if the encoding is in the weights rather than the weakened input, the same contact-order and flexibility signals should be recoverable at the default MSA depth with stronger attenuation (λ below the 0.70–0.85 grid tested here); the paper leaves behavior at default depth untested, and that experiment would localize the signal.
  • My extension: because perturbation sensitivity matches flexibility patterns but not magnitudes, the probe could serve as a simulation-free per-residue flexibility prior for fold classes well represented in the training corpus — worth benchmarking against crystallographic B-factors and NMR order parameters on a wider panel than the three proteins studied.
  • My extension: the five divergent α-synuclein landscapes generate a concrete, testable hypothesis the author leaves open — that model-specific basins correspond to real functional sub-ensembles (extended helical states resembling the membrane-bound form, compact states resembling aggregation-prone forms); mapping each model's basin to a functional context would distinguish structured extrapolation
  • My extension: the block-2 anomaly (catastrophic collapse when exactly the first three blocks are perturbed, absent at depths 2 and 4) and the KaiB depth-7 dip indicate the 48 Evoformer blocks are not functionally interchangeable; per-block and per-head perturbation profiles could map a coarse-to-fine division of labor and reveal where the network compensates.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper introduces Scaled Gaussian Convolution (SGC), a deterministic perturbation of AlphaFold2/OpenFold Evoformer weights, and argues that the resulting structural responses reveal a conformational landscape encoded in the learned weights. Using ubiquitin as the main test system, the author reports that native contacts break in an experimentally established stability ordering (G2 before G3, G1 last), that per-residue perturbation sensitivity correlates with microsecond MD RMSF beyond geometric baselines, that the pooled perturbed ensemble reproduces the L-shaped folding funnel topology of millisecond MD, and that matched-power noise controls do not produce such structured responses. The results are further supported by five-model replication and by application to KaiB (convergent non-recovery of the alternative fold) and α-synuclein (structured inter-model disagreement). The paper is candid about several limitations, especially the use of reduced MSA depth (64×64) in all primary analyses.

Significance. If the central claim holds, the paper is significant: it would show that a structure-prediction model trained only on static targets nonetheless encodes a readable, physically meaningful ordering of conformational organization, and it introduces a generally applicable perturbation-based interpretability protocol. The external validation strategy is a genuine strength: the primary comparisons are against DE Shaw MD trajectories, Went/Jackson and Sosnick folding experiments, and five separately trained OpenFold replicas, with no parameter fitted to reproduce the target orderings. The noise controls and MSA-depth comparisons are also thoughtful. However, the significance is conditional on the reduced-MSA issue described below; the paper itself labels default-depth behavior an open question, which tempers the abstract's claims.

major comments (4)
  1. [§2.3, §4.4, Supplementary S11] All quantitative headline results (contact-loss ordering in §3.5, RMSF correlations in §3.6, landscape topology in §3.4) are produced at MSA depth 64×64. The paper notes that at 128×128 atomization falls from 3.0% to 0.2% and that default 512×5,120 behavior is untested. This is load-bearing: the central claim is that the Evoformer weights encode a conformational landscape, but the readout is only demonstrated in a regime where the coevolutionary input is deliberately weakened. The author's cross-depth controls (32/64/128) mitigate but do not eliminate the concern that the structured response is an artifact of an under-constrained transformer rather than a faithful readout of the weights. A control at default MSA depth—possibly with larger perturbation strength—or a demonstration that the 64×64 response is a monotone amplification of a default-depth signal is required to support the abstr
  2. [§3.1, §3.5] The depth axis conflates 'more blocks corrupted' with 'more early-stage computation warped.' The paper explicitly invokes perturbed/unperturbed block mismatch to explain tail recovery (Section 3.1), yet interprets the depth-resolved contact-loss ordering (Section 3.5) as differential encoding robustness. Cumulative forward perturbation changes both the number of corrupted blocks and the location of the perturbed/unperturbed boundary; the strong forward/reverse asymmetry reported in Section 3.1 shows that boundary effects are large. Without single-block or windowed perturbation controls, the G2-before-G3 ordering could reflect the specific position of the boundary rather than a stability ordering encoded in the weights. The paper defers such controls to future work, but this gap is central to the interpretation of Section 3.5.
  3. [§3.6, Table 2] The flexibility correlation claim is pattern-level, not magnitude-level, in the regime where it is strong. At depths 4–24 the partial r|WCN is 0.66–0.79, but Lin's CCC is about 0.08 and perturbation sensitivity is about 18-fold smaller than MD RMSF. At depths 27–31 the magnitudes converge (CCC=0.66) but the structures are partially unfolded and outside the native basin sampled by the 300 K MD. The paper states this tradeoff honestly, but the conclusion that 'Evoformer weights carry a residue-level constraint signal correlated with physical flexibility' rests on a reduced-MSA, pattern-only correlation in the shallow regime and an unfolded-ensemble magnitude agreement in the deep regime. A more direct test—e.g., comparing SGC sensitivity on native-basin structures at higher MSA depth—would strengthen the causal claim.
  4. [§3.11, §4.2] The α-synuclein 'shared-core' prediction is not sharply falsifiable as stated. The shared-core region in (Rg, Ree) space is defined by the overlap of the five AF2 models and the pooled MD ensemble, and the prediction is that any future experimental ensemble with sufficient sampling will occupy this region. Because the region is partly defined by the MD reference, the prediction risks circularity. To make it testable, the region should be defined from the five-model intersection alone (without MD), and the prediction should specify a quantitative occupancy threshold (e.g., a measurable fraction of experimental frames) and a sampling criterion.
minor comments (5)
  1. [Abstract / §2.3] The abstract states that 'the conformational organization visible under perturbation... emerged as a byproduct' without noting that all quantitative support comes from the reduced-MSA regime and that default-MSA behavior is untested. A one-sentence qualification would align the abstract with Section 4.4.
  2. [§3.1 / §3.2] Depth-1 is described as an anomaly in Section 3.8 but is not consistently excluded in the summary metrics; the exclusions (depths ≤5 versus depths ≤3) should be stated in each figure/table caption.
  3. [Supplementary S8 / main text §3.10] Cross-references to 'main Fig. 13' in the supplementary gallery should be updated to the correct main-text figure numbers (the KaiB landscape appears to be Figure 14).
  4. [Code and data availability] The statement that code and processed tables are 'available from the author on request' is insufficient for a study of this scale and for the reproducibility claims made in the text. At minimum, the code, YAML manifests, and processed metric tables should be deposited in a public repository.
  5. [§3.7] The sentence 'all structures at depths 2 and 4–24 maintain a fully connected polypeptide chain' is contradicted by the 18% of runs with broken bonds reported in the same section. Clarify that the 18% refers to deeper depths and the phrase 'excluding... anomalies' should be explicit here as well.

Circularity Check

1 steps flagged · score 3.0 of 10

Minor definitional circularity in the α-synuclein shared-core prediction; central results rest on external validation.

  1. self definitional [Section 3.11, Fig. 17C; restated in Section 4.2]
    "The shared core, where all models agree and MD also finds density, is consistent with constraints shared across training data and physical simulation. ... We predict that this region will contain the experimentally occupied ensemble for any future characterization of α-synuclein's disordered state with sufficient sampling."

    The 'shared core' is defined as the overlap of the five AF2 models with the pooled MD density (Fig. 17C). The prediction that future experimental ensembles will fall inside it is therefore anchored to the MD reference used to draw the region; the prediction inherits reference occupancy by construction rather than being an independent forecast from the weights. This does not affect the main ubiquitin/KaiB validations, which use external MD, folding experiments, and five OpenFold replicas.

full rationale

The load-bearing quantitative claims—contact-loss ordering vs. Went/Jackson and Sosnick experiments, RMSF correlations vs. DE Shaw equilibrium MD, KaiB non-switching across five models, and the matched-power noise controls—are checked against external references without fitting parameters to reproduce the target orderings. There are no self-citations by the author (all cited perturbation methods and AlphaFold results are by other groups), no imported uniqueness theorems, and no ansatz smuggled in via the author's prior work. The reduced-MSA regime is an acknowledged limitation ('Behaviour at default MSA depth... remains an open question') and a correctness risk, not a circular step, since the ordering and topology are preserved across MSA depths 32/64/128 and are separately validated against physical references. The only definitional short-circuit is the α-synuclein 'most falsifiable prediction': the shared-core region is selected as the AF2-through-MD overlap, so predicting that future experimental ensembles will occupy it is a restatement of the boundary choice. This is a minor, non-load-bearing element; the central derivation chain is largely self-contained with external support.

Assumptions & free parameters 8 free parameters · 9 assumptions · 3 invented entities

The paper's scientific payload is external benchmarking of a deterministic weight-perturbation probe, with the probe's axes (λ, σ, d), the MSA depth, the block-2 exclusion, and the contact-group definitions all being hand-chosen or post-hoc selections that shape the results. No parameter was fitted to a biological target, but the G1/G2/G3 grouping imports the expected ordering from the same literature used as the benchmark, and the α-synuclein prediction inherits MD occupancy by construction.

free parameters (8)
  • λ (scaling factor) = 0.75 primary; grid 0.70–0.85
    Dominant perturbation axis (>99.5% of power at σ=0.30). Primary value chosen post-hoc from the response map to maximize the 'dynamics zone' (RMSD 1.5–5.0 Å), not fitted to any biological target.
  • σ (Gaussian width) = 0.30 primary; grid 0.00–0.30
    Blur width; contributes 0.43% of perturbation power and is 87.5% parallel to pure scaling at the operating point (Table S4). Swept to map the transition boundary; structural role confined to a narrow λ window.
  • d (perturbation depth) = 1–48, cumulative forward
    Defines the axis along which the 'unfolding pathway' and contact-loss ordering are measured; confounded with number of early blocks corrupted (mismatch with unperturbed blocks 24–47).
  • MSA depth = 64×64 clusters/extras (default 512×5,120)
    Chosen because 'reduced MSA depth... make the model more sensitive to perturbation'; all quantitative claims live in this regime; sensitivity falls monotonically with depth and default-MSA behavior is untested (Sec. 2.3, S11).
  • Block-2 anomaly exclusion (depths ≤5) = excluded from max-RMSD/min-Q, landscape, and transition analyses
    Deterministic, unexplained depth-3 collapse (plus depth-1 excursions) removed from summaries; documented but not mechanistically integrated (Sec. 3.1, 3.8).
  • Contact cutoffs for Q-factor groups = heavy-atom 4.5 Å; Cα–Cα 12 Å; seq. sep. ≥3
    Hand-chosen; the 12 Å Cα cutoff is generous and was selected to match secondary-structure contacts in the folding literature, i.e., tuned to the benchmark.
  • G1/G2/G3 contact-group memberships = 93/45/112 contacts
    Defined from refs [7–9] — the same literature that supplies the expected stability ordering used as the benchmark; grouping and expectation are not independent.
  • Primary RMSF reference (CHARMM22*/TIP3P, 154.6 μs) = one of six equilibrium force fields
    Primary flexibility benchmark; cross-force-field consistency shown (5 native-state FFs cluster at r>0.7), so this choice is weakly load-bearing.
assumptions (9)
  • domain assumption OpenFold v2.2.0 model_1_ptm parameter set (93M) stands in for 'AlphaFold2's' learned weights
    The abstract/title attribute the results to AlphaFold2; experiments run OpenFold's retrained parameters, which share architecture but not DeepMind's exact training run (Sec. 2.3, ref [6]).
  • domain assumption At MSA depth 64×64, structural response is attributable to the learned weights rather than the reduced coevolutionary input
    Explicit design logic in Sec. 2.3; the opposite attribution (floppiness from input under-supply) is not excluded.
  • domain assumption 33-structure recycling trajectories can be treated as a 'conformational ensemble'
    The author explicitly disclaims thermodynamic/Boltzmann/kinetic readings (Sec. 2 terminology), so the assumption is bounded but still load-bearing for the RMSF correlation.
  • standard math Kabsch-aligned Cα RMSD × Q-factor projection preserves the salient folding-funnel topology for both AF2 and MD
    Used for the 72.6% territory-overlap claim; projection and affine normalization choices affect the comparison (Sec. 2.5).
  • standard math WCN partial correlation isolates non-trivial flexibility signal
    The headline flexibility finding (partial r|WCN 0.66–0.79) depends on this control being adequate; GNM baseline suggests the control works.
  • domain assumption DE Shaw MD trajectories [9,12] are valid external reference for topology and per-residue flexibility
    External ground truth; a99SB-UCB shows extensive unfolding and is treated as an out-of-basin control — the reference pool itself is heterogeneous.
  • domain assumption The experimental stability ordering of ubiquitin (refs [7,8]) is the correct benchmark for contact-loss order
    One protein, one literature line; the paper acknowledges the ordering reflects denaturation, not the kinetic pathway.
  • ad hoc to paper The block-2 collapse is an 'early-interface compensation failure' and safe to exclude from summaries
    Labeled hypothetical by the author (Sec. 3.1); exclusion is a modeling choice, not a demonstrated mechanism.
  • ad hoc to paper Perturbed/unperturbed block mismatch explains tail recovery but not mid-depth ordering
    The author invokes the mismatch to explain depths 44–48 recovery (Sec. 3.1) without controlling for the same mismatch along the rest of the depth axis.
invented entities (3)
  • Scaled Gaussian Convolution (SGC) protocol
    purpose: Deterministic weight perturbation (scale × blur) intended as a 'spectroscopic' probe of encoded conformational structure
    A method, not an entity; no falsifiable handle independent of the paper.
  • 'Encoded conformational landscape' in the Evoformer weights
    purpose: Postulated representational object claimed to contain stability orderings, flexibility patterns, and funnel topology beyond single-structure predictions
    The one external handle offered — the α-synuclein shared-core prediction — is weakly falsifiable because the region was defined partly by existing MD occupancy (Sec. 3.11).
  • 'Neural spectroscopy' framing
    purpose: Names the research program of reading trained networks via controlled weight perturbation
    Terminological framing; evidence burden carried by SGC results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural spectroscopy of AlphaFold2 reveals encoded protein conformational landscapes." pith.science (2026). https://pith.science/paper/JZFI4PLT

@misc{pith2026260716087,
  author       = {Pith},
  title        = {Pith review of: Neural spectroscopy of AlphaFold2 reveals encoded protein conformational landscapes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JZFI4PLT}},
  note         = {Machine review of arXiv:2607.16087}
}
read the original abstract

AlphaFold2's 93 million parameters, shaped by the evolutionary record of protein structure encoded in the Protein Data Bank and in sequence alignments, are conventionally treated only as machinery for converting sequence to structure. We propose they are also a scientific object that can be analyzed directly: a learned encoding of protein conformational organization that can be probed and characterized. By smoothing the Evoformer's weight tensors with a Gaussian convolution and scaling the result, we show that the trained model produces physically structured conformational landscapes. Under perturbation, ubiquitin's native contacts break in the order established by decades of folding experiments. For KaiB, five independently trained models agree that the alternative fold is not recovered under perturbation. For alpha-synuclein, five models produce five different but coherent landscapes, mapping where the training signal has determined the representation and where it has not. Matched-power noise controls confirm that random corruption of equal magnitude produces debris, not conformations. The model learned to predict static structures; the conformational organization visible under perturbation was not an explicit training target, suggesting it emerged as a byproduct of that objective. AlphaFold2's weights appear to encode structural constraints, shaped by evolutionary and structural training data, that extend beyond what unperturbed inference reveals. We call the approach of reading them neural spectroscopy, and Scaled Gaussian Convolution one such protocol.

Figures

Figures reproduced from arXiv: 2607.16087 by the authors.

Figure 1
Figure 1. Scaled Gaussian convolution (SGC) perturbs Evoformer weights to probe encoded [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. SGC perturbation induces a tunable structural transition. [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Conformational landscape of SGC-perturbed ubiquitin. [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Parameter response map of the SGC parameter space. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: Sigma and lambda provide independent control over the structural response. [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: The dynamics zone is narrow and tunable. [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Recycling trajectories on the conformational landscape. [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: The conformational landscape mirrors the MD folding funnel. (A) [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Perturbation depth traces a physically ordered unfolding pathway. (A–D) [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: Perturbation sensitivity tracks physical flexibility. (A) [PITH_FULL_IMAGE:figures/full_fig_p023_10.png]
Figure 11
Figure 11. Figure 11: Noise controls reveal that weight-aligned coherence is necessary for controllable [PITH_FULL_IMAGE:figures/full_fig_p025_11.png]
Figure 12
Figure 12. Figure 12: The encoding is model-independent. (A–E) [PITH_FULL_IMAGE:figures/full_fig_p027_12.png]
Figure 13
Figure 13. Figure 13: SGC shifts pair-activation magnitudes along a single structured axis. (A) [PITH_FULL_IMAGE:figures/full_fig_p029_13.png]
Figure 14
Figure 14. Figure 14: KaiB: the conformational landscape is one-dimensional denaturation, not fold [PITH_FULL_IMAGE:figures/full_fig_p031_14.png]
Figure 15
Figure 15. Figure 15: Perturbation depth controls denaturation extent, not switching direction. [PITH_FULL_IMAGE:figures/full_fig_p032_15.png]
Figure 16
Figure 16. Figure 16: α-synuclein: five models, five landscapes. (A–E) Per-model conformational landscapes (− ln ρ density in Rg–Ree space), pooling all recycles across 3 perturbation conditions, 48 depths, and 3 seeds (∼28,000 frames per model). Each model produces a distinct landscape to…
Figure 17
Figure 17. Figure 17: α-synuclein: AF2 ensemble covers the MD-sampled core. (A) Pooled AF2 landscape across all 5 models (141,375 frames) in Rg–Ree space. (B) DE Shaw explicit-solvent MD ensemble pooled across eight force-field variants (∼273 µs total; Supplementary Table S1) in the same c…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

40 extracted references · 16 canonical work pages

  1. [1]

    Jumper, J., Evans, R., Pritzel, A. et al. Highly accurate protein structure prediction with AlphaFold.Nature596, 583–589 (2021). doi:10.1038/s41586-021-03819-2

    doi:10.1038/s41586-021-03819-2 2021

  2. [2]

    & Perez, A

    Chang, L. & Perez, A. Rapid estimation of protein folding pathways from sequence alone using AlphaFold2.Nature Communications17, 170 (2026). doi:10.1038/s41467-025-66870-x

    work page doi:10.1038/s41467-025-66870-x 2026

  3. [3]

    & Wallner, B

    Kalakoti, Y. & Wallner, B. AFsample2 predicts multiple conformations and ensembles with AlphaFold2.Communications Biology8, 373 (2025). doi:10.1038/s42003-025-07791-9

    work page doi:10.1038/s42003-025-07791-9 2025

  4. [4]

    Wayment-Steele, H.K., Ojoawo, A., Otten, R. et al. Predicting multiple conformations via sequence clustering and AlphaFold2.Nature625, 832–839 (2024). doi:10.1038/s41586-023- 06832-9

    doi:10.1038/s41586-023- 2024

  5. [5]

    Taneja, I., Llanos, M.A., Fernández-Quintero, M.L. et al. AlphaFold-RandomWalk and AlphaFold-Ensemble: sampling alternative protein conformations with perturbed versions of AlphaFold.J. Chem. Inf. Model.66, 152–166 (2026). doi:10.1021/acs.jcim.5c02004

    work page doi:10.1021/acs.jcim.5c02004 2026

  6. [6]

    Ahdritz, G., Bouatta, N., Floristean, C. et al. OpenFold: retraining AlphaFold2 yields new insights into its learning mechanisms and capacity for generalization.Nature Methods21, 1514–1524 (2024). doi:10.1038/s41592-024-02272-z

    doi:10.1038/s41592-024-02272-z 2024

  7. [7]

    & Jackson, S.E

    Went, H.M. & Jackson, S.E. Ubiquitin folds through a highly polarized transition state. Protein Eng. Des. Sel.18, 229–237 (2005). doi:10.1093/protein/gzi025

    work page doi:10.1093/protein/gzi025 2005

  8. [8]

    & Krantz, B.A

    Sosnick, T.R., Dothager, R.S. & Krantz, B.A. Differences in the folding transition state of ubiquitin indicated byϕ and ψ analyses.Proc. Natl. Acad. Sci. USA101, 17377–17382 (2004). doi:10.1073/pnas.0407683101

    work page doi:10.1073/pnas.0407683101 2004

Show all 40 references
  1. [9]

    & Shaw, D.E

    Piana, S., Lindorff-Larsen, K. & Shaw, D.E. Atomic-level description of ubiquitin folding. Proc. Natl. Acad. Sci. USA110, 5915–5920 (2013). doi:10.1073/pnas.1218321110

    2013 doi

  2. [10]

    & Cook, W.J

    Vijay-Kumar, S., Bugg, C.E. & Cook, W.J. Structure of ubiquitin refined at 1.8 Å resolution. J. Mol. Biol.194, 531–544 (1987). doi:10.1016/0022-2836(87)90679-6

    1987 doi

  3. [11]

    & Egli, M

    Pattanayek, R., Williams, D.R., Pattanayek, S., Mori, T., Johnson, C.H., Stewart, P.L. & Egli, M. Structural model of the circadian clock KaiB–KaiC complex and mechanism for modulation of KaiC phosphorylation.EMBO J.27, 1767–1778 (2008). doi:10.1038/emboj.2008.104

    2008 doi

  4. [12]

    & Shaw, D.E

    Robustelli, P., Piana, S. & Shaw, D.E. Developing a molecular dynamics force field for both folded and disordered protein states.Proc. Natl. Acad. Sci. USA115, E4758–E4766 (2018). doi:10.1073/pnas.1800690115

    2018 doi

  5. [13]

    & Skolnick, J

    Zhang, Y. & Skolnick, J. TM-align: a protein structure alignment algorithm based on the TM-score.Nucleic Acids Res.33, 2302–2309 (2005). doi:10.1093/nar/gki524

    2005 doi

  6. [14]

    43 Neural spectroscopy of AlphaFold2

    Vankerschaver, J.tmtools: Python bindings for TM-align.https://github.com/jvkersch/ tmtools(2022). 43 Neural spectroscopy of AlphaFold2

    2022

  7. [15]

    A solution for the best rotation to relate two sets of vectors.Acta Crystallogr

    Kabsch, W. A solution for the best rotation to relate two sets of vectors.Acta Crystallogr. A32, 922–923 (1976). doi:10.1107/S0567739476001873

    1976 doi

  8. [16]

    & Beckstein, O

    Michaud-Agrawal, N., Denning, E.J., Woolf, T.B. & Beckstein, O. MDAnalysis: A toolkit for the analysis of molecular dynamics simulations.J. Comput. Chem.32, 2319–2327 (2011). doi:10.1002/jcc.21787

    2011 doi

  9. [17]

    & Eaton, W.A

    Best, R.B., Hummer, G. & Eaton, W.A. Native contacts determine protein folding mech- anisms in atomistic simulations.Proc. Natl. Acad. Sci. USA110, 17874–17879 (2013). doi:10.1073/pnas.1311599110

    2013 doi

  10. [18]

    McGibbon, R.T., Beauchamp, K.A., Harrigan, M.P. et al. MDTraj: A modern open library for the analysis of molecular dynamics trajectories.Biophys. J.109, 1528–1532 (2015). doi:10.1016/j.bpj.2015.08.015

    2015 doi

  11. [19]

    Tseng, R., Goularte, N.F., Chavan, A. et al. Structural basis of the day-night transition in a bacterial circadian clock.Science355, 1174–1180 (2017). doi:10.1126/science.aag2516

    2017 doi

  12. [20]

    & Porter, L.L

    Chakravarty, D. & Porter, L.L. AlphaFold2 fails to predict protein fold switching.Protein Sci.31, e4353 (2022). doi:10.1002/pro.4353

    2022 doi

  13. [21]

    Chakravarty, D., Porter, L.L. et al. AlphaFold predictions of fold-switched conforma- tions are driven by structure memorization.Nature Communications15, 7296 (2024). doi:10.1038/s41467-024-51801-z

    2024 doi

  14. [22]

    & Pappu, R.V

    Ruff, K.M. & Pappu, R.V. AlphaFold and implications for intrinsically disordered proteins. J. Mol. Biol.433, 167208 (2021). doi:10.1016/j.jmb.2021.167208

    2021

  15. [23]

    Elhage, N., Nanda, N., Olsson, C. et al. A mathematical framework for transformer cir- cuits.Transformer Circuits Thread, Anthropic (2021). https://transformer-circuits. pub/2021/framework/index.html

    2021

  16. [24]

    & Wolynes, P.G

    Bryngelson, J.D., Onuchic, J.N., Socci, N.D. & Wolynes, P.G. Funnels, pathways, and the energy landscape of protein folding: A synthesis.Proteins21, 167–195 (1995). doi:10.1002/prot.340210302

    1995 doi

  17. [25]

    Olsson, C., Elhage, N., Nanda, N. et al. In-context learning and induction heads.Trans- former Circuits Thread, Anthropic (2022). https://transformer-circuits.pub/2022/ in-context-learning-and-induction-heads/index.html

    2022

  18. [26]

    Abramson, J., Adler, J., Dunger, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3.Nature630, 493–500 (2024). doi:10.1038/s41586-024-07487-w

    2024 doi

  19. [27]

    & Jaakkola, T

    Jing, B., Berger, B. & Jaakkola, T. AlphaFold Meets Flow Matching for Generating Protein Ensembles.Proceedings of the 41st International Conference on Machine Learning (ICML) (2024). arXiv:2402.04845

    2024 arXiv

  20. [28]

    Lin, Z., Akin, H., Rao, R. et al. Evolutionary-scale prediction of atomic-level protein structure with a language model.Science379, 1123–1130 (2023). doi:10.1126/science.ade2574 44 Neural spectroscopy of AlphaFold2

    2023 doi

  21. [29]

    & Meiler, J

    Del Alamo, D., Sala, D., Mchaourab, H.S. & Meiler, J. Sampling alternative conforma- tional states of transporters and receptors with AlphaFold2.eLife11, e75751 (2022). doi:10.7554/eLife.75751

    2022 doi

  22. [30]

    & Belinkov, Y

    Meng, K., Bau, D., Andonian, A. & Belinkov, Y. Locating and editing factual associations in GPT.Advances in Neural Information Processing Systems35(2022). arXiv:2202.05262

    2022 arXiv

  23. [31]

    & Kern, D

    Wayment-Steele, H.K., Otten, R., Pitsawong, W., Ojoawo, A.M., Glaser, A., Calderone, L.A. & Kern, D. The conformational landscape of fold-switcher KaiB is tuned to the circadian rhythm timescale.Proc. Natl. Acad. Sci. USA121, e2412293121 (2024). doi:10.1073/pnas.2412293121

    2024 doi

  24. [32]

    & Ivankov, D.N

    Pak, M.A., Markhieva, K.A., Novikova, M.S., Petrov, D.S., Vorobyev, I.S., Maksimova, E.S., Kondrashov, F.A. & Ivankov, D.N. Using AlphaFold to predict the impact of single mutations on protein stability and function.PLOS ONE18, e0282689 (2023). doi:10.1371/journal.pone.0282689

    2023 doi

  25. [33]

    & Tlusty, T

    McBride, J.M., Polev, K., Abdirasulov, A., Reinharz, V., Grzybowski, B.A. & Tlusty, T. AlphaFold2 can predict single-mutation effects.Phys. Rev. Lett.131, 218401 (2023). doi:10.1103/PhysRevLett.131.218401

    2023 doi

  26. [34]

    & Porter, L.L

    Schafer, J.W. & Porter, L.L. AlphaFold2’s training set powers its predictions of some fold-switched conformations.Protein Sci.34, e70105 (2025). doi:10.1002/pro.70105

    2025 doi

  27. [35]

    & Nussbaum, R.L

    Ulmer, T.S., Bax, A., Cole, N.B. & Nussbaum, R.L. Structure and dynamics of micelle-bound humanα-synuclein.J. Biol. Chem.280, 9595–9603 (2005). doi:10.1074/jbc.M411805200

    2005 doi

  28. [36]

    Tuttle, M.D., Comellas, G., Nieuwkoop, A.J. et al. Solid-state NMR structure of a pathogenic fibril of full-length humanα-synuclein.Nat. Struct. Mol. Biol.23, 409–415 (2016). doi:10.1038/nsmb.3194

    2016 doi

  29. [37]

    Varadi, M., Anyango, S., Deshpande, M. et al. AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models.Nucleic Acids Research50, D439–D444 (2022). doi:10.1093/nar/gkab1061

    2022 doi

  30. [38]

    & Moult, J

    Kryshtafovych, A., Schwede, T., Topf, M., Fidelis, K. & Moult, J. Critical assessment of methods of protein structure prediction (CASP)—Round XV.Proteins91, 1539–1549 (2023). doi:10.1002/prot.26617

    2023 doi

  31. [39]

    & Pierce, B.G

    Yin, R., Feng, B.Y., Varshney, A. & Pierce, B.G. Benchmarking AlphaFold for pro- tein complex modeling reveals accuracy determinants.Protein Science31, e4379 (2022). doi:10.1002/pro.4379

    2022 doi

  32. [40]

    cumulative forward

    Bakan, A., Meireles, L.M. & Bahar, I. ProDy: Protein Dynamics Inferred from Theory and Experiments.Bioinformatics27, 1575–1577 (2011). doi:10.1093/bioinformatics/btr168 45 Supplementary Information Neural spectroscopy of AlphaFold2 reveals encoded protein conformational landsc...

    2011 doi

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.