The abstract is read at the publisher; the summary is JClub's.
MRCAT achieved scientific completion—meeting all nine prespecified MR procedural requirements—for 26 executable Mendelian randomization queries and correctly handled all four boundary cases.
- Why it matters: MR reliability depends on coordinated, assumption-consistent procedures and transparent interpretation, yet AI automation can finish computationally without meeting scientific standards.
- What they did: The multi-agent workflow uses dependency-aware planning, skill-guided execution, expert-informed review, STROBE-MR–guided traceable reporting, and iterative Writer-Judge checking across 30 natural-language queries.
- The result: In six tasks where expert-curated analyses did not support causal effects, MRCAT matched expert numerical results more closely and produced consistent interpretations in all six cases.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.