A safety net, not a substitute. Every result must be confirmed by standard laboratory antimicrobial susceptibility testing and reviewed by a qualified healthcare or laboratory professional.
Why we picked Challenge Six: Genome Firewall, how we built it, and how we measure whether it worked.
Antimicrobial resistance is projected to cause up to 10 million deaths per year by 2050. Clinicians increasingly have access to whole-genome sequencing but no calibrated, auditable way to turn that data into an antibiotic-response signal.
Challenge Six sits exactly at that gap. It is a defensive research tool that reads an assembled genome and emits a decision receipt: likely to work, likely to fail, or a deliberate no-call. We picked it because the safety framing (abstain rather than guess) matches how we already think about ML in high-stakes settings.
Three of us, one rule: don't ship what you can't defend (quod erat demonstrandum).
We built a lab-report aesthetic (monochrome paper, highlighter accents, monospace for the technical data) so every screen reads like something you could staple into a notebook. The Analysis page presents Analyze, Evidence & Trace, and Model Validation.
Held-out AUROC ≥ 0.90 per drug and ECE ≤ 0.05 after calibration on unseen genetic groups.
Under 90 s end-to-end per genome on commodity CPU, with an honest no-call layer.
An auditable receipt for every decision, zero silent failures, scope explicit on the UI.
Decision support alongside phenotypic AST, flagging high-risk isolates for confirmation faster.
Population-scale AMR trend detection from sequenced isolates with calibrated uncertainty.
Prioritize candidate compounds against genotypes predicted to escape existing drugs.
External references follow the recommendations in the challenge description.