· 8 min read
Introducing Nextflow Frontier
A recursive closed-loop discovery system for the unsolved problems in life sciences. On BioML-bench it has already scored better than 92% of leaderboard entries.
Notes from the frontier
How we build a closed-loop discovery system, and what we learn running it with our partners.
11 min
Drug discovery
Ranking molecules for a cancer drug target
Nextflow Frontier ranked 144 molecules for EGFR binding with two small chemistry models, scoring 0.887 on PR-AUC against 0.866 for a billion-parameter network.
11 min
Biomedical imaging
Forecasting lung decline from a single CT scan
Nextflow Frontier forecast lung-function decline in pulmonary fibrosis from one CT scan, missing later measurements by 170 mL on average against 228 mL.
12 min
Protein engineering
Ranking protein edits by how they change stability
Nextflow Frontier ranked length-changing protein edits with a model about 80 times smaller than ESM-1v, scoring 0.953 against 0.940 on 46 variants.
11 min
Single-cell omics
Naming kidney cell types from their RNA
Nextflow Frontier classified 5,278 kidney cells from a held-out donor into 13 cell types, scoring 0.977 on weighted F1 against 0.963 for the XGBoost baseline.
5 min
Benchmarks
How we benchmark Nextflow Frontier
Why we ran Nextflow Frontier on BioML-bench: to show it respects its contract, to test our infrastructure under load, and to train our forward deployed scientists.
