Research Publications and Lab Newsletters
Institutions on AlkhemYst-Ai.cloud share findings, workflows, and experiment results with the research community.
Nineteen known fungi, and what decides which ones a run can see
AlkhemYst-Ai
A fungal mock community control rebuilt from its public reads: no readout sees all nineteen fungi, ITS1 and ITS2 from the same library together see every one, and each fungus lost is lost to the length of the read or the fit of a primer
October 1, 2026
Two algae that were in the reads all along
AlkhemYst-Ai
A 2016 comparison of 18S primers rebuilt from its seven mock communities: V4 loses two haptophytes and V8-V9 does not, as the paper found, and V4's own reverse reads hold both algae, lost in the forward reads after a GC-rich stretch
October 9, 2026
A sample of known answers: rebuilding a classic mouse gut study from its raw reads
AlkhemYst-Ai
The 2012 mouse weaning study behind the most used amplicon teaching set, rebuilt from its raw reads: the mock community came back exact, and three of the paper's four claims held within the one animal the public data holds.
October 2, 2026
What it takes to check a paper, when the authors have already done the hard part
AlkhemYst-Ai
A 2024 pharmacology paper from Tromso, rebuilt in an afternoon because its authors had already done the hard part: they deposited what was needed to check them.
September 22, 2026
The result held. The data grain changed what we could test.
AlkhemYst-Ai
The 2023 Nature paper on European freshwater biodiversity, led from Senckenberg with NORCE in Bergen, rebuilt outside its original environment. The driver analysis reproduced. The temporal breakpoint could not be tested at all.
September 7, 2026
When a warning flag is not a diagnosis
AlkhemYst-Ai
A UiT cohort study on school dropout, re-examined. A variable associated with a hard outcome is not automatically a predictor of it once the rest of a student situation is in the model.
August 26, 2026
When a held-out split is not reproducible: an Alchemyst-Crucible audit
AlkhemYst-Ai
Kvasir-SEG is a real contribution from Simula Research Laboratory. Its 80/10/10 split cannot be reproduced from what was published, so no later baseline is comparable to the original.
August 26, 2026
I reproduced a 2018 methylation paper. The authors were right, and their evidence could not show it.
AlkhemYst-Ai
A 2018 DNA methylation study from Oslo University Hospital, rebuilt from its deposited data. The finding held. The evidence offered for it could not have shown it either way.
August 14, 2026