Decision benchmarking of cryo-EM re-refinement

In a prespecified 3.0–3.6 Å cryo-EM re-refinement cohort (167 entries; 131 sequence groups), highest-correlation selection increased chi1 disagreement by 2.4° (95% CI 2.1–2.7°) versus retaining the deposited model.

• Why it matters: Map fit selection is commonly used during re-refinement, but its effect relative to doing nothing is unclear, especially when judged against higher-resolution side-chain agreement.
• What they did: The study benchmarked selection rules against a no-operation baseline using packed-side-chain agreement with higher-resolution references, including a half-map withheld selection test.
• The result: Half-map withheld selection increased chi1 disagreement by 2.29° (1.70–2.97°; 99 entries), and in 53 low/high-resolution pairs the losses were 3.43° and 2.35°, enabling Primum to benchmark refiners and selection rules.

AI summary of the abstract · may contain errors · not medical advice · read the paper.