Review papers with raw data transparency

What the paper did (succinct; with citation)

Wu et al. collected 778 PDB spike trimers (UniProt P0DTC2) solved by cryo‑EM/X‑ray, aligned them to PDB 8DLW, computed center‑of‑mass (COM) motions for four S1 domains and defined two orientational parameters (α, β) using Cα anchors (S514, C361) to quantify RBD hinge rotation and an effective rotation around a constructed hinge point P3; β correlates with d_RBD and α captures twist during rising, producing a reproducible RBD‑rising pathway and two principal crown‑opening routes (uud > udu in antibody‑bound data). Wu et al. 2025 (preprint, methods & results)

Strengths (evidence-focused)

• Very large structural sample (n=778) pooling public PDB data — increases statistical signal for geometric metrics derived from static structures Wu et al. dataset
• Clear, reproducible geometric definitions (α, β) with precision checks (σ ≈2.4°) and alternative-anchor checks reported — helps standardize comparisons across labs and PDB entries Precision checks (α,β)
• Integrates interpretation with orthogonal experimental literature (FRET states, MD free‑energy barriers) and provides concrete mapping between structural intermediates and smFRET signals smFRET contextualization

Key limitations & potential biases (must consider before using α/β)

1. PDB sampling bias: static structures deposited in PDB reflect experimental choices (constructs, stabilizing mutations, detergents, antibody complexes) — many conformations are enriched or depleted relative to native virions. The authors acknowledge this and remove 9 structures with altered loops near C361, but full sampling bias remains an important caveat Sampling limitation noted
2. Two‑atom anchors simplify a complex motion: α/β rely on S514 and C361 Cα positions; while precision tests showed σ≈2.4°, these two atoms are a coarse proxy and can miss local loop rearrangements or glycan-mediated motions (glycans modulate opening energetics in MD studies) — so α/β quantify a dominant rigid‑body mode but not full conformational complexity MD: multi‑pathway opening
3. Alignment/coordinate-frame dependence: α and β values depend on alignment to reference 8DLW and on which S2 helices are used; although authors align using S2 helices and test precision, small misalignments propagate into α/β for near‑threshold cases (β ~20–35°) — threshold choice (β=35°) is pragmatic but somewhat arbitrary.
4. Interpretational leap to dynamics: static snapshots cannot directly measure kinetics; mapping to FRET and MD is plausible and useful, but claims about energy barriers and sequential order (uud vs udu) need direct time‑resolved validation (smFRET, HDX-MS, or long MD/Gō‑Martini sampling) rather than PDB cooccurrence counts alone Mapping to dynamic experiments

Detailed methodological critique (concise points)

• Selection criteria: inclusion: full/near‑full protomers <4 Å resolution — this is defensible; however, resolution alone does not guarantee accurate loop placement (important for C361 and RBM loops) — authors trimmed structures missing key Cαs from COM calculations (reasonable). Structure selection & alignment
• Angle definitions: α and β are geometrically well defined and tested (σ ~2.4°). Authors also tested group‑Cα alternatives with slightly higher σ — good practice; but users should re-calc α/β when using different reference frames or mutated constructs (e.g., heavily glycosylated or stabilized spikes). Angle robustness checks
• Bound‑state classification (4 Å criterion): operationally clear but sensitive — some antibody contacts are mediated by glycans or long sidechains so a 4 Å non‑hydrogen atom cutoff may misclassify weak binders; authors defined bound‑spike as carrying ≥1 bound‑RBD which is pragmatic for population-level trends.

What is likely correct (high confidence) vs speculative (low confidence)

• High confidence: α and β are reproducible geometric descriptors extractable from PDB models and useful to quantitatively compare RBD orientations across structures; authors demonstrate internal precision and show linear β–d_RBD correlation (β–d_RBD correlation).
• Moderate confidence: antibodies broaden β distributions and shift populations toward larger β values — paper shows clear differences between bound/unbound histograms; effect magnitude may depend on antibody class and epitope (steric vs allosteric) and on construct differences among deposited PDBs (Antibody effect on β).
• Lower confidence / speculative: the proposed dominant crown‑opening route (uud > udu) in vivo and the mechanistic electrostatic push/pull explanation inferred from single structures — these are plausible hypotheses but require time‑resolved experiments (smFRET, long MD, single‑particle cryo‑EM of native virions, or hydrogen‑deuterium exchange) to confirm sequence and kinetics (Need for smFRET/time resolved validation).

Concrete recommendations (short‑term actionable)

1. Reproduce α/β calculation code and release it: the paper gives methods but not code; authors should publish scripts to compute α/β from PDB (alignment → extract Cα coordinates → compute P3/P4/P5 → α, β) for community validation.
2. Compare α/β computed on curated virion‑like cryo‑ET datasets (in situ spikes) to test PDB selection bias (Ke et al. 2020 style) — if in‑situ spike β profiles differ, that signals experimental artefacts in deposited constructs In‑situ spike heterogeneity paper.
3. Time‑resolved validation: pair smFRET experiments where donor/acceptor placements map to the same d_RBD/β geometry used in this paper; correlate observed FRET states to α/β bins to test the intermediate‑state assignment (paper already suggests some mapping) smFRET validation.
4. MD / coarse‑grained simulations: run targeted MD or optimized Gō‑Martini runs seeded from identified intermediate PDBs to test transitions and energetic barriers (HFC‑refined CG maps can increase sampling of metastable states) GōMartini contact‑map improvements.

Falsification conditions (what would disprove the paper's main claims)

• If time‑resolved smFRET or long unbiased MD shows that RBD opening occurs via routes inconsistent with α/β pathway (e.g., multiple orthogonal pathways with no neck at β≈16–35°), the proposed general RBD‑rising pathway would be falsified.
• If in‑situ cryo‑ET of native virions yields β distributions markedly different from PDB-derived ones (e.g., synchronized triple opening common), the sequential one‑by‑one opening claim would need revision.
• If experimental perturbations (mutations or antibodies) expected, per α/β classification, to increase β do not change FRET kinetics or receptor binding in biochemical assays, the functional mapping would be falsified.

Suggested short reproducible analysis you can run now

1) Compute α/β for a curated set of 100 PDBs (release code + Jupyter notebook). 2) Plot β distributions for antibody‑bound vs unbound subsets and run Kolmogorov‑Smirnov test to quantify distribution shift. Wu provides alignment coordinates and domain residues to replicate steps (Reproducibility resources (paper)).

Selected citations used in this review

• Wu et al. 2025 (primary)
• smFRET review
• MD free‑energy context
• GōMartini HFC improvements
• In‑situ virion cryo‑ET
• ChimeraX reference