Structure prediction
Atomic models of proteins and full biomolecular complexes — ligands, nucleic acids, and modified residues — in one framework.
Public milestones · 2020–2026
AI for Biology
Mapping molecules to medicine — how machine learning is reshaping structure, design, and discovery across the living world.
A curated showcase of verified public progress. Not medical advice.
01 — Why it matters
From sequence to structure to function — and then to the clinic.
Biology is chemistry at planetary scale. Proteins fold, genomes encode, cells decide. For decades, mapping that chain — sequence → structure → function → therapy — was painfully slow. AI is compressing each step: predicting shapes, designing new proteins and molecules, reading multi-omics atlases, and automating the lab loop that turns hypotheses into experiments.
StructureShape of life’s machines
FunctionWhat those shapes do
DesignNew proteins & drugs
ImpactMedicines, crops, materials
02 — Timeline
Milestones that moved the field
Verified public themes. Wording is careful where claims are still early or company-reported.
-
2020–2024
AlphaFold 2 era & Nobel Prize
AlphaFold 2 (DeepMind, 2020–2021) cracked high-accuracy protein structure prediction. The companion AlphaFold Protein Structure Database, with EMBL-EBI, released predictions for over ~200 million protein sequences. In October 2024, Demis Hassabis and John Jumper shared half of the Nobel Prize in Chemistry for protein structure prediction; David Baker received the other half for computational protein design.
Nobel Prize press release · AFDB / EMBL-EBI
-
May 2024
AlphaFold 3 — beyond proteins alone
Published in Nature (8 May 2024), AlphaFold 3 predicts joint structures of complexes including proteins, DNA, RNA, small-molecule ligands, ions, and modified residues. DeepMind and Isomorphic Labs reported substantially improved accuracy versus prior specialized tools across many interaction types — including protein–ligand and protein–nucleic acid cases highlighted in the paper.
Abramson et al., Nature 2024 · doi:10.1038/s41586-024-07487-w
-
Ongoing
Protein language models — the ESM family
Meta’s Evolutionary Scale Modeling (ESM) lineage treats amino-acid sequences like language: large transformers learn evolutionary grammar, enabling structure and function inference at massive scale from sequence alone. ESMFold and related models complement structure predictors and open high-throughput annotation of metagenomic “dark matter.”
Meta AI / ESM publications & open models
-
Design wave
Generative protein & molecule design
Diffusion and graph models turned prediction into invention. Tools such as RFdiffusion (backbone generation) and ProteinMPNN (sequence design) — alongside generative chemistry platforms — let researchers propose novel binders, enzymes, and drug-like molecules, then filter them in silico before wet-lab synthesis.
Baker lab & community open-source methods
-
2025–2026
Clinical: Insilico’s rentosertib enters Phase III
Rentosertib (ISM001-055 / INS018_055), an AI-originated small molecule from Insilico Medicine targeting TNIK for idiopathic pulmonary fibrosis (IPF), has been reported as the first generative-AI-originated drug to advance into late-stage (Phase III) trials. Company and press statements in September 2026 announced first-patient dosing in GENESIS-IPF-3 (NCT07687459). The candidate remains investigational — no regulatory approval, and this site makes no efficacy claims.
Insilico / PR Newswire · ClinicalTrials.gov NCT07687459
-
2025–2026
Isomorphic Labs · IsoDDE
Alphabet’s Isomorphic Labs published the IsoDDE (Drug Design Engine) technical report (February 2026), describing a unified computational system that aims to go beyond AlphaFold 3 for structure, ligand binding, affinity, and antibody–antigen tasks. Public comments have targeted first-in-human trials for an AI-designed candidate around end of 2026 — timelines and candidates remain company-reported and subject to change.
isomorphiclabs.com · Pharma Letter / Semafor coverage
-
Broader stack
Genomics, single-cell, imaging & the automated lab
Parallel waves span foundation models for DNA/RNA, single-cell atlases, AI microscopy and phenotyping, and closed-loop lab automation — shortening the cycle from hypothesis to measured result across academia and industry.
Genomics FMs Single-cell Cell imaging Lab automation
03 — Frontiers
Where the frontier is moving
Drug design
In silico pocket finding, affinity estimation, and generative chemistry feeding preclinical pipelines and early clinical programs.
Protein design
De novo binders, enzymes, and nanomaterials designed by diffusion and sequence models — then validated at the bench.
Genomics & multi-omics
Foundation models and atlases that connect genotype, transcriptome, and proteome to phenotype across tissues and species.
Imaging & phenotyping
AI that reads cells and tissues at scale — from microscopy segmentation to high-content screens and digital pathology.
Lab automation
Robotic wet labs and closed-loop experiment planners that turn model proposals into measured data overnight.
04 — By the numbers
Scale you can cite
Only publicly reported figures. Soft wording where counts evolve.
~200M+
Predicted structures in the AlphaFold Protein Structure Database (DeepMind + EMBL-EBI; literature also cites ~214 M sequences covered).
2M+
People reported by the Nobel Committee as having used AlphaFold 2, across ~190 countries (2024 Chemistry Prize materials).
2024
Nobel Prize in Chemistry shared by Baker (design) and Hassabis & Jumper (structure prediction).
Phase III
Company-reported status of Insilico’s rentosertib for IPF — first generative-AI-originated small molecule described as entering late-stage trials (investigational).
05 — Sources
Primary references
- Abramson et al. — Accurate structure prediction of biomolecular interactions with AlphaFold 3
- Google DeepMind — AlphaFold overview & research
- Nobel Prize in Chemistry 2024 — Press release
- AlphaFold Protein Structure Database
- Isomorphic Labs — IsoDDE Drug Design Engine
- Insilico Medicine — GENESIS-IPF-3 first-patient dosing (rentosertib)
- ClinicalTrials.gov NCT07687459 — Rentosertib Phase 3 IPF study
This showcase summarizes public reporting for a scientific audience. It is not medical advice, an endorsement of any company or drug, or a claim of clinical efficacy. Trial status and company timelines can change.