Microcin H47 cancer-fighting peptide
An experimental peptide studied for its ability to attack cancer cells; not an approved drug.
A researcher, an agent, or an algorithm wrote down the sequence and picked a target to hit.
An AI model like OpenFold3 or AlphaFold built a 3D structure and scored how well it fits the binding site.
A second contributor repeated the computation on their own hardware and the scores matched.
Literature-extracted sequence peptide — synthesized for bioassay as documented in linked reference(s)
Fork this card to add platform evidence →
Activity measured in linked reference(s) — IC50/MIC/cytotoxicity data
Fork this card to add platform evidence →
Research directions for this peptide, selected from the current sources — hypotheses you can explore and model. None of it is proven yet; tap any one to see the full thinking.
Has anyone identified the actual molecule this peptide acts on inside a cancer cell, or is 'anticancer' just shorthand for 'kills cells in a dish'?
This matters for anyone trying to improve or combine this peptide. Microcin H47 is a known antibacterial peptide that works by sneaking through bacterial iron-uptake channels and blocking the cell's ATP synthase, so its 'anticancer' tag may be an unvalidated screening result with no human target behind it.
Could this antibacterial peptide reduce the gut bacteria linked to colorectal cancer, offering an indirect benefit rather than directly killing tumor cells?
If true, it could be delivered to the gut rather than injected, a simpler path. One caveat: this peptide's natural killing range is fairly narrow, so whether it actually hits cancer-linked bacteria like Fusobacterium would need to be tested before any claim of benefit.
Could a trimmed core of this peptide keep its activity while being cheaper to make, or are the floppy ends actually required?
A shorter peptide would be far cheaper to produce. A caveat to test first: in natural Microcin H47 the serine-rich C-terminal region carries an essential chemical modification needed for activity, so simply cutting that end could abolish function rather than preserve it.
▸full evidence table1 metrics
| metric | value | tool |
|---|---|---|
| ranking score | 0.6037186980247498 | boltz-2 |
▸3-letter notation
▸recipeboltz-2 2.2.1
| parameter | value |
|---|---|
| model | boltz-2 2.2.1 |
| weights | — |
| hardware | vast_v100_32gb |
| mlx version | — |
| python | — |
| random seed | 1 |
| msa strategy | none_monomer |
| runtime | — |
| predicted by | — |
| predicted at | 2026-05-23 |
▸citationbibtex
@peptide{pep05201,
sequence = {GGAPATSANAAGAAAIVGALAGIPGGPLGVVVGAVSAGLTTAIGSTVGSGSASSSAGGGS},
target = {anticancer},
author = {peptidemodel},
year = {2026},
status = {bioassayed}
}