Finding new medicines for fungal mycetoma, a disease that often ends in amputation
- Good it would do
- 2 / 5
- Can AI help
- 2 / 5
- How neglected
- 4 / 5
- Can we check it
- 4 / 5
Why it's on hold
Parked: the disease is severe and neglected and the MycetOS data are genuinely open, but Ersilia's public model already does the proposed computation (the lab reproduced its 0.886 cross-validated score with a twenty-line random forest, against 0.849 for plain similarity search), and what limits the one project doing this work is physical and organisational: about 20-30 molecules synthesised a year, empty compound vials, a seasonal larval assay, only 45 public in vivo results, and two unanswered public offers of computational help. The outside critic's fatal rating rested on a misread malaria paper and was withdrawn by the skeptic.
Why it's stuck
Mycetoma attracts very little research. A bibliometric study found only about 400 to 500 papers on it in the 21 years from 1999 to 2020.
What we'd try
MycetOS publishes all of its screening results openly under a Creative Commons licence, including an early screen in which 215 of 800 compounds stopped the fungus growing. Compute could train models on this open data and screen very large virtual compound libraries for new active chemical series with good drug-like properties.
We'd look again if
Any one of: (1) a named partner at MycetOS, Erasmus MC or DNDi owns the relationship, answers within a week and commits specific synthesis or larval-assay slots to the lab's output; (2) PENEGRAIN (2024-2029) or UNRAVELMy publishes enough grain-penetration measurements to model the step that decides whether a compound is a medicine; (3) a funded route exists to buy or synthesise the molecules on a ranked list. Then run the pre-registered benchmark in the dossier: precision at a 20-30 compound budget under time and scaffold splits, against the published eos4f95 model, nearest-neighbour similarity and a one-line azole rule.
Think we got it wrong?
Ruled-out problems can come back. Tell us what we missed.
Members can read the full record for this problem, check its sources and comment. Open the member view →