Review papers with raw data transparency

Key takeaways (visual-first, then evidence)

• Accurate framing of opportunity: The authors correctly identify computational bottlenecks in multi-omics and drug discovery and map quantum primitives (Grover, VQE, QAOA, annealing) to those problems (Nassir et al. (2025) review mapping).
• Correctly emphasises hardware limits: decoherence, qubit counts, and error correction are central constraints preventing immediate clinical application — authors describe QEC approaches (surface codes) and mitigation (ZNE) accurately (Preskill (2018) on NISQ era).
• Balance of hype vs evidence: review cites many proof-of-concept QML and small-scale drug-discovery studies but finds no clinical validation — the paper properly notes absence of clinical datasets and states "No datasets were generated or analysed" (Nassir et al. — data statement).

Detailed critique (strengths, blindspots, and priority experiments)

Strengths

1. Comprehensive: covers hardware, algorithms, software stacks, and biomedical use-cases with abundant citations (Nassir et al. — comprehensive refs).
2. Translational focus: explicitly proposes hybrid workflows and digital-twin concepts that are actionable research directions rather than pure theory (Nassir et al. — translational roadmap).

Key blindspots and limitations

1. Evidence grade: Many cited QML/image/diagnostic claims improve accuracy by small margins on curated datasets — but these are not clinical trials, and small-sample or dataset-selection bias is likely; the review acknowledges this but does not quantify effect-size heterogeneity across studies (Nassir et al. — diagnostic claims).
2. Optimism on timelines: timeline figures and language imply steady maturity, but they underweight non-technical barriers (regulatory validation, clinical workflow integration) that often dominate translational timelines for diagnostics and therapeutics (Nassir et al. — timeline optimism).
3. Reproducibility & data encoding: the paper notes encoding heterogeneous clinical and multi-omics data into qubit registers is unresolved but stops short of recommending concrete standards or reference pipelines for benchmarking — a missed opportunity to accelerate reproducible comparisons (Nassir et al. — encoding limitations).

High-priority experiments to falsify/validate claims

1. Head-to-head benchmark: multi-site, multi-cohort retrospective test where a quantum-enhanced hybrid pipeline and an optimized classical pipeline are compared on the same preprocessed multi-omics diagnostic task with pre-registered metrics (AUC, sensitivity, calibration). If quantum method fails to exceed classical after reasonable tuning, claim of near-term advantage is falsified (Preskill (2018) on benchmarking).
2. Drug-discovery prospective validation: run a hybrid quantum-classical virtual-screen workflow prospectively (ranked compounds), then pass top hits to blinded experimental binding assays and orthogonal cellular readouts; evaluate hit rate and enrichment vs classical in identical compute budget. Failure to improve hit enrichment is falsifying for practical advantage claims (Nassir et al. — KRAS example).

Concrete recommendations (research & translational priorities)

• Establish open benchmark datasets and standardized quantum-data-encoding conventions for multi-omics tasks (so comparisons are reproducible).
• Create pre-registered hybrid pipeline challenges (diagnostics, biomarker selection, virtual screening) with compute- and energy-budget limits.
• Prioritise interpretable QML models with calibration and uncertainty estimates for clinical decision-support.
• Make hardware-agnostic middleware and reproducible containerized quantum-classical workflows (e.g., PennyLane + common preprocessing) publicly available.

Minimal evidence map (selected citations from the review)

1. Nassir et al., npj Genomic Medicine (2025)
2. Preskill (2018) on NISQ