Life & medical sciences · OECD FoS 1.6 · China MoE 0710 生物学
Life sciences (structural & molecular biology)
生命科学（结构与分子生物学）
L3Systemic transformation ↑ Rising Score 3.45 · Last reviewed 2026-10-22 · Next review 2026-12-20
BIGGEST BREAKTHROUGH 2025-11-05 AI leapField leap
Atomically accurate de novo design of antibodies with RFdiffusion (Nature) ↗
Antibodies were the hardest de novo design target; cryo-EM confirms accuracy
Why this level
L3 (close to L4): AlphaFold made structure prediction the default step and won the 2024 chemistry Nobel; RFdiffusion reported atomically accurate de novo antibodies (Nature 2025); Evo 2 (Nature, Mar 2026) models and designs genomes across life. Missing: independent adoption surveys and curriculum evidence.
S = 0.30·method + 0.35·discovery + 0.25·adoption + 0.10·education (education scores 0 when there is no evidence; if empty, the other weights are renormalised)
Sources
Composite 2.5–3.49 (or ≥3.5 with any of method/discovery/adoption below 3): AI is standard infrastructure with multiple independent breakthroughs. All levels →
Research directions
Protein structure predictionDe novo protein designProtein language modelsBiomolecular complexes
Papers
- Programmable design of functional proteins from natural language
AI: trained on 1.7B protein–text pairs, makes natural language the main interface for protein design (open weights). Field: intended function observed in all four tasks; an unoptimised H-protein raised CO₂-fixation titre 1.7-fold
2026-10-21 · Nature
- Atomic functional site scaffolding in biopolymers with RFdiffusion3
AI: diffusion generates proteins at all-atom level around catalytic sites; Field: with two Nature papers on de novo proteases, enzyme design moves from binders to catalysis
2026-10-15 · Science
- Computational design of metalloproteases
AI: an AI pipeline designs new metalloproteases that cleave peptide bonds; Field: proteases were long out of reach for de novo design
2026-10-14 · Nature
- AI-designed antibodies with Germinal
Antibody design becomes a reusable pipeline
2026-08-06 · Nature Methods
-
AI leapField leap
Atomically accurate de novo design of antibodies with RFdiffusionAntibodies were the hardest de novo design target; cryo-EM confirms accuracy
2025-11-05 · Nature
- Simulating 500 million years of evolution with a language model
Generates a distant new fluorescent protein
2025-02-21 · Science
- Accurate structure prediction of biomolecular interactions with AlphaFold 3
Extends to complexes with nucleic acids and ligands
2024-05-08 · Nature
-
Field leap
De novo design of protein structure and function with RFdiffusionDiffusion-based de novo binder design with high success rates
2023-07-11 · Nature
- Evolutionary-scale prediction of atomic-level protein structure with a language model
Extends prediction to metagenomic “dark matter”
2023-03-17 · Science
-
AI leapField leap
Highly accurate protein structure prediction with AlphaFoldNear-experimental accuracy at CASP14; the 50-year folding problem
2021-07-15 · Nature
- Facilitating Valid Statistical Inference in Biomedical Image Synthesis
Tests whether generative-AI synthetic MRI images support valid downstream statistical and biological inference, not just visual similarity.
2026-10-16 · bioRxiv · Xiong, J., Bao, S., Liu, Q., Lau, K.
- Homology-aware cross-validation strategies for generalization assessment in RNA secondary structure prediction
Proposes homology-aware cross-validation to avoid leakage-driven over-optimism in deep-learning RNA secondary structure prediction.
2026-10-16 · bioRxiv · Bugnon, L., Kulemeyer, G., Gerard, M., Di Persia, L.
- Evolutionary Informed Translation from Protein to Nucleic Acid Sequence
PINATA uses phylogenetic alignments to translate proteins into more natural, species-specific nucleic-acid sequences.
2026-10-15 · arXiv · Anna Hart, Mohammed El-Kebir, Huimin Zhao, Heng Ji
- ExpertTwins: Auditable Long Horizon Scientific Panels through Grounded Expertise
Defines each seat of a multi-agent scientific panel by an inspectable evidence policy so agents do not simply converge.
2026-10-15 · bioRxiv · Foroughi pour, A., Shirinifard, A.
- scMTG reconstructs single-cell temporal dynamics with Markov transition generators
scMTG reconstructs cell-state transitions from destructive single-cell time-series data.
2026-10-14 · bioRxiv · Cui, X., Wang, H., Gao, Z., Jiang, R.
- Joint Variable Selection for Omic Biomarkers in Time-to-Event Data
Joint variable selection linking time-to-diagnosis in biobank EHRs to multi-omic measurements.
2026-10-14 · bioRxiv · Bajzik, J., Depope, A., Zolfimoselo, Y., Sharipov, A.
- DNA-LLM2Vec: Architecture-Aware Task-Conditioned Adaptation of Decoder DNA Foundation Models
Adapts directional decoder DNA foundation models into bidirectional, task-conditioned representations for genomic tasks.
2026-10-14 · bioRxiv · Liang, Y., Zhai, K., Gan, Y., Li, S.
- Do LLM Proposers Improve Budgeted Search for Ionizable Lipids?
LEAF tests, at matched budget over 30,044 ionisable lipids, whether LLM proposers beat classical search.
2026-10-13 · bioRxiv · Singh, H., SINGH, R. K., Krishna, C. S., Srivastava, S. P.
- From arXiv, bioRxiv, medRxiv and OpenAlex (last ~30 days); summaries written by AIMap from the abstract.
Milestones 13· 5 leaps
Big glowing nodes are leaps: AI leap cool · Field leap warm · double ring = both
-
Model
Pinal: a 16B-parameter foundation model that designs functional proteins from natural-language descriptions, validated in four experimental tasks (Nature)AI: trained on 1.7B protein–text pairs, makes natural language the main interface for protein design (open weights). Field: intended function observed in all four tasks; an unoptimised H-protein raised CO₂-fixation titre 1.7-fold
Nature -
Paper
RFdiffusion3: atomic functional-site scaffolding in biopolymers (Science)AI: diffusion generates proteins at all-atom level around catalytic sites; Field: with two Nature papers on de novo proteases, enzyme design moves from binders to catalysis
Science -
Paper
Computational design of metalloproteases (Nature)AI: an AI pipeline designs new metalloproteases that cleave peptide bonds; Field: proteases were long out of reach for de novo design
Nature -
Paper
Germinal AI-designed antibodies with validated nanomolar binders for four targets (Nature Methods)Antibody design becomes a reusable pipeline
Nature Methods -
System
Isomorphic Labs unveils IsoDDE, dubbed “an AlphaFold 4” (not public)Vendor claim, undisclosed
Nature NewsVendor claim -
PaperAI leapField leap
Atomically accurate de novo design of antibodies with RFdiffusion (Nature)Antibodies were the hardest de novo design target; cryo-EM confirms accuracy
Nature -
Paper
ESM3 designs a new fluorescent protein (Science)Generates a distant new fluorescent protein
Science -
PrizeAI leapField leap
AlphaFold and protein design win the Chemistry NobelNobel prize for AI structure prediction and design
Nobel Prize -
Paper
AlphaFold 3 predicts complexes with DNA/RNA/ligands (Nature)Extends to complexes with nucleic acids and ligands
Nature -
PaperField leap
RFdiffusion de novo protein design (Nature)Diffusion-based de novo binder design with high success rates
Nature -
Paper
ESMFold: language-model structure prediction for 600M+ metagenomic proteins (Science)Extends prediction to metagenomic “dark matter”
Science -
DatasetField leap
AlphaFold DB expands to 200M+ structures200M+ structures, nearly all known proteins
Google DeepMind -
PaperAI leapField leap
AlphaFold 2 paper published (Nature)Near-experimental accuracy at CASP14; the 50-year folding problem
Nature
No milestones match this filter.
More papers & repos 12
| Type | Title | Date |
|---|---|---|
| Paper | Computational design of metalloproteases | 2026-10-14 |
| Paper | AI-designed antibodies with Germinal | 2026-08-06 |
| Paper | Atomically accurate de novo design of antibodies with RFdiffusion | 2025-11-05 |
| Paper | Simulating 500 million years of evolution with a language model | 2025-02-21 |
| Paper | Accurate structure prediction of biomolecular interactions with AlphaFold 3 | 2024-05-08 |
| Paper | De novo design of protein structure and function with RFdiffusion | 2023-07-11 |
| Paper | Evolutionary-scale prediction of atomic-level protein structure with a language model | 2023-03-17 |
| Paper | Highly accurate protein structure prediction with AlphaFold | 2021-07-15 |
| Repo | google-deepmind/alphafold3 To be verified | — |
| Repo | RosettaCommons/RFdiffusion To be verified | — |
| Repo | evolutionaryscale/esm To be verified | — |
| Repo | jwohlwend/boltz To be verified | — |
Latest news 12
- CADS: Measuring the Incremental Contribution of LLM Proposers in Sequential Biological ScreensarXiv q-bioenPaperPreprint
- Disentanglement with Biological Constraints: A Theory of Functional Cell TypesarXiv q-bioenPaperPreprint
- Hybrid modeling for improved structure prediction of large protein complexesNature Computational Scienceen
- Researchers use AI to design proteins that cut up other proteinsArs Technica AIen
- Chai Discovery signs GSK deal to make biologics with AIEndpoints Newsen
- Designing enzymes for new-to-nature chemistry and non-natural substrates with AlphaProtein NovobioRxiv (synthetic biology)enPreprint