The Evidence Layer for Scientific AI

Your phrase “Best Evidence: Enomoto Key Papers Summaries” appears to be a label, but the summaries themselves (text/DOIs/results) are not included. Without the primary evidence content, I cannot:

• verify which claims correspond to which experiments/figures (claim-to-evidence mapping)
• check study design quality (controls, blinding, sample sizes, model system comparability)
• detect overreach (correlation vs causation, external validity limits)
• produce raw-data-aligned visuals (only possible when numeric results are present)

Next step: paste the “Enomoto Key Papers Summaries” text here, or provide the DOI/PMID list + the exact summary bullets you want assessed.

1. Claim extraction: split each summary into atomic claims (mechanism, association, effect size, directionality, boundaries).
2. Evidence binding: for each claim, identify the specific primary endpoint(s) and figure/table references (e.g., viability assay, binding Kd, RNA-seq DE genes, electrophysiology traces).
3. Study design critique: evaluate controls, comparators, randomization/blinding (if stated), dose/time/model choices, and whether outcomes are readouts or mechanisms.
4. Bias & reproducibility screening: assess publication/reporting risk (selective reporting patterns), missing negative results, and whether independent replication is present in the paper set.
5. Cross-paper consistency: check if effect directions match across models/species/assays; flag contradictions and plausible reasons (technique differences, batch effects, genotype background).
6. Quantification & uncertainty: extract numeric effect sizes when available (means, SD/SEM, CIs, n) and show uncertainty in plots.
7. Falsification map: identify what observations would disprove the strongest claims.

• For each paper: DOI (preferred) or PMID, plus the summary text you want evaluated.
• If summaries already reference figures/tables, include those identifiers.
• If numeric results are mentioned (effect sizes, p-values, fold-changes), include them verbatim.

AI Scientist mode

If you paste the summaries/DOIs, this agent will iteratively extract claims, locate evidence passages (within the provided text set), and produce critical visuals + a “what would falsify this?” section.