Cost and open-science framing
What this project actually cost
Resource | Cost |
AlphaFold Server (Q6VY07, 8 AF3 jobs) | 0 € |
ChEMBL database | 0 € (open) |
AutoDock Vina, RDKit, Meeko, fpocket, OpenMM | 0 € (open source) |
ESMFold API | 0 € (free for research) |
OSF hosting | 0 € (free for public projects) |
Compute (2 CPU, ~2 days continuous) | ~50 € (electricity) |
Personal workstation amortization | ~5 k€ over 3 years (~5 €/day pro-rated) |
Total | ~50 € + 5 days of work |
What the wet-lab equivalent would cost
For the same final conclusion (3 selective binders + pharmacophore + 5 controls):
Step | Wet-lab method | Cost | Time |
Structure (WT) | X-ray crystallography | 80–150 k€ | 1.5–3 yr |
Structure (R203W) | mutagenesis + re-crystallography | 30–50 k€ | 8–12 mo |
Pocket detection | manual expert analysis | ~0 € | (limited) |
Screening 1,397 drugs | SPR + cellular phenotype | 60–200 k€ | 6–12 mo |
Robustness controls | independent triplicates | ~50 k€ | 2–3 mo |
Pharmacophore (36 analogues) | purchase + DSF/SPR | ~15 k€ | 1–2 mo |
Pocket residue ID | co-crystallography + cross-link MS | ~50 k€ | 6–12 mo |
Random control panel | DSF on 7 random | ~2 k€ | 2 wk |
Publication-grade validation | orthogonal SPR/ITC | ~25 k€ | 4–6 mo |
Total | 310–540 k€ | 3–5 yr |
Ratios
Criterion | Wet-lab | Computational | Ratio |
Direct cost | 310–540 k€ | ~50 € | 6,000–10,000× cheaper |
Time | 3–5 yr | 5 days | 200–365× faster |
Failure risk per step | 40–60% (structural steps) | ~10% (transparent) | Lower-risk |
Reproducibility | Costly replicates | Reproducible at zero cost | Infinite |
Pivot speed | Slow | Quasi-immediate | Infinite |
Caveats
This comparison should not be misread as "computational replaces wet-lab":
- Wet-lab validation remains essential. The retrospective cohort + SPR binding assay will cost ~50–80 k€ but provide the experimental confirmation that this project cannot.
- AlphaFold's contribution is excluded from the wet-lab cost. Including it would push the wet-lab equivalent even higher.
- The retrospective cohort is the highest ROI next step — data already collected, weeks not years, zero direct cost.
Why open data matters
All ligand structures, AF3 inputs, docking results, control panel data, scripts, and reports are on OSF (DOI: 10.17605/OSF.IO/ED7SP). Any researcher can re-run the screening with different parameters, test additional compounds against the same pocket, run the deferred MD on GPU, or cite the work for credit-traceable use.
We believe rare-disease drug repurposing should be open-source by default. The patient communities cannot afford otherwise.