Molecular Brain — 97% of authors couldn't produce raw data; 21 papers withdrawn

When the editor of Molecular Brain asked 41 authors of submitted manuscripts to produce the raw data behind their results, 97% could not, and 21 of the papers were withdrawn (S-0049).

The 41-paper sample is small but the methodology is direct: papers were already in the editorial pipeline at a peer-reviewed journal; the editor asked authors to produce the underlying data; nearly every author failed.

The 21-paper withdrawal rate is the operational consequence: in the editorial workflow, when raw data is asked for and cannot be produced, papers are withdrawn rather than published. The case is a microcosm of what happens when verification is enforced at scale: the publication funnel narrows substantially, because the underlying data infrastructure does not support a verification-based regime.

The architectural reading: in the steady state, the gap between "results produced" and "raw data inspectable" is the 97% figure here. Most papers fall through the gap by default; the 21 withdrawals here became visible only because the editor asked.