Life & medical sciences · OECD FoS 1.6 · China MoE 071007 遗传学
Genomics & genetics
基因组学与遗传学
L3Systemic transformation ↑ Rising Score 2.80 · Last reviewed 2026-10-22 · Next review 2026-12-20
BIGGEST BREAKTHROUGH 2026-03-04 AI leapField · major
Evo 2: 40B-parameter genome model with 1 Mb context across all domains of life (Nature) ↗
Trained on 9.3T bases; predicts variant effects and designs validated sequences
Why this level
Deep-learning variant calling (DeepVariant) is in standard pipelines; AlphaMissense scores 71M missense variants (Science 2023); AlphaGenome reads 1 Mb of DNA to predict thousands of tracks (Nature, Jan 2026); genome LMs like Evo generate functional sequences.
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
Variant effect predictionRegulatory genomicsDNA language modelsSingle-cell foundation models
Papers
- Predicting genome-wide functional constraints with GPN-Star
AI: a phylogeny-aware design lets a compact gLM beat classical evolutionary models; Field: better pathogenic/GWAS variant prioritisation and rare-variant tests, extended to five model organisms
2026-09-09 · Nature
-
AI leap
Genome modelling and design across all domains of life with Evo 2Trained on 9.3T bases; predicts variant effects and designs validated sequences
2026-03-04 · Nature
- Advancing regulatory variant effect prediction with AlphaGenome
1 Mb input, thousands of functional tracks
2026-01-28 · Nature
- Sequence modeling and design from molecular to genome scale with Evo
First DNA LM generating functional sequences
2024-11-15 · Science
- Accurate proteome-wide missense variant effect prediction with AlphaMissense
Pathogenicity for 71M missense variants
2023-09-22 · Science
- Effective gene expression prediction from sequence by integrating long-range interactions
Extends sequence-to-function context to 100 kb
2021-10-04 · Nature Methods
- A genomic foundation model for interpreting human genetic diversity
Pretrained on 4.784B haplotype sequences from 63,809 WGS individuals; ranks first in 18 of 19 primary comparisons.
2026-10-21 · bioRxiv · Q. Xu, Z. Wu, M. J. Li
- Harness-Driven Agentic Operation of Optical Pooled Screen Analysis
Brieflow-auto, a Claude Code harness, lets an agent run multi-day optical pooled screen analyses end to end.
2026-10-15 · arXiv · Matteo Di Bernardo, Claire Peterson, Ege Topkoc, Matt Might
- Panoramic modelling of single-cell transcriptomes at scale with CellNet
CellNet models a broader range of single-cell transcriptomic information to characterise cellular diversity.
2026-10-15 · arXiv · Xiaoqing Lian, Pei Liu, Jiahui Tan, Luoda Tan
- Decoding super-enhancer regulatory grammar and pathogenic gain-of-function variants directly from DNA sequence
Borzoi-SE, built on the Borzoi genomic transformer, identifies cell-type super-enhancers and their grammar from sequence alone.
2026-10-14 · bioRxiv · Liang, Y., Xu, X., Mao, Y., Zheng, Z.
- Deep learning-based virtual perturbation reveals increasing heterogeneity of gene regulatory responses with aging
Deep-learning virtual knockouts (DKO) show gene-regulatory responses become more heterogeneous with ageing.
2026-10-13 · arXiv · Yongjian He, Julia Hanna Haddad Guez, Xu-Wen Wang, Orr Levy
- BioMod: Biological Module Tokenization for Single-Cell Foundation Modeling
Tokenises cells by biological gene modules rather than single genes for single-cell foundation models.
2026-10-13 · bioRxiv · Meng, L., Zhu, J., Han, B., Zang, Y.
- CellChain: Benchmarking Cross-Scale Evidence Use for Perturbation Reasoning
A benchmark for whether LLMs link molecular mechanisms, cellular responses and outcomes when reasoning about perturbations.
2026-10-12 · arXiv · Xue Xia, Yue Wang, Shuizhou Chen, Lang Yu
- From arXiv, bioRxiv, medRxiv and OpenAlex (last ~30 days); summaries written by AIMap from the abstract.
Milestones 8· 1 leaps
Big glowing nodes are leaps: AI leap cool · Field leap warm · double ring = both
-
Model
hEVO: a human genomic foundation model trained on 4.784B haplotype sequences from 63,809 genomes, ranking first in 18 of 19 primary comparisons (bioRxiv)AI: pretrained on WGS-reconstructed haplotypes rather than a reference genome. Field: leads Evo 2, NT, GPN etc. in 18/19 comparisons on annotation, variant interpretation and evolution (self-reported preprint)
bioRxivPreprint -
Model
GPN-Star: phylogeny-informed genomic language model for genome-wide functional constraints (Nature)AI: a phylogeny-aware design lets a compact gLM beat classical evolutionary models; Field: better pathogenic/GWAS variant prioritisation and rare-variant tests, extended to five model organisms
Nature -
PaperAI leap
Evo 2: 40B-parameter genome model with 1 Mb context across all domains of life (Nature)Trained on 9.3T bases; predicts variant effects and designs validated sequences
Nature -
Paper
Evo: sequence modelling from molecular to genome scale (Science)First DNA LM generating functional sequences
Science -
Paper
AlphaMissense predicts 71M missense variants (Science)Pathogenicity for 71M missense variants
Science -
Paper
Enformer predicts gene expression from sequence with long-range interactions (Nature Methods)Extends sequence-to-function context to 100 kb
Nature Methods -
System
DeepVariant deep-learning variant caller (Nature Biotechnology)Variant calling as image classification; entered standard pipelines
Nature Biotechnology
No milestones match this filter.
More papers & repos 9
| Type | Title | Date |
|---|---|---|
| Paper | Predicting genome-wide functional constraints with GPN-Star | 2026-09-09 |
| Paper | Genome modelling and design across all domains of life with Evo 2 | 2026-03-04 |
| Paper | Advancing regulatory variant effect prediction with AlphaGenome | 2026-01-28 |
| Paper | Sequence modeling and design from molecular to genome scale with Evo | 2024-11-15 |
| Paper | Accurate proteome-wide missense variant effect prediction with AlphaMissense | 2023-09-22 |
| Paper | Effective gene expression prediction from sequence by integrating long-range interactions | 2021-10-04 |
| Repo | google/deepvariant To be verified | — |
| Repo | google-deepmind/alphagenome To be verified | — |
| Repo | ArcInstitute/evo2 To be verified | — |
Latest news 12
- CPTCs Drive Somatic-Visceral Communication via the Wnt Axis in Somatic Mechanotherapy: A Single-Cell Deep Learning StudyarXiv q-bio.BMenPaperPreprint
- RNAGenScape: Property-Guided, Optimized Generation of mRNA Sequences with Manifold Langevin DynamicsarXiv q-bio.BMenPaperPreprint
- gpuGEM: GPU-based linear optimisation of genome-scale metabolic modelsarXiv q-bioenPaperPreprint
- Evo2HiC: a multimodal foundation model for integrative analysis of genome sequence and architecturebioRxiv (bioinformatics)enPreprint