AlphaFold hits ‘next level’: the AI tool now includes protein pairing

## Summary
For the first time, the AlphaFold protein-structure database will include predictions of complexes of proteins — with the addition of 1.7 million ‘homodimers’ comprising two interacting strands of the same molecule. AlphaFold is five years old — these charts show how it revolutionized science Such proteins were already included in the database as individual ‘monomers’ but their entries tell only part of their story. “We thought, ‘can we bring the AlphaFold database to the next level, where we can include a lot of complex predictions across the tree of life?’” says Martin Steinegger, a computational biologist at Seoul National University in South Korea, who was part of the effort. Complex interactions To make predictions for even small complexes of two proteins was a crucial challenge, says Steinegger. “It is quite a different beast than monomer predictions.” Protein-complex predictions are exceedingly intensive computationally, so a consortium — including Steinegger’s lab, EMBL-EBI, Google DeepMind and chipmaker NVIDIA in Santa Clara, California — was formed to take on the challenge. Article PubMed Google Scholar Download references Reprints and permissions Related Articles The huge protein database that spawned AlphaFold and biology’s AI revolution AlphaFold is running out of data — so drug firms are building their own version Chemistry Nobel goes to developers of AlphaFold AI that predicts protein structures AlphaFold touted as next big thing for drug discovery — but is it?

## Article Content
Email
Bluesky
Facebook
LinkedIn
Reddit
Whatsapp
X
AlphaFold is now capable of predicting homodimeric complexes, including those formed by the transcription elongation factor Eaf, whose N‑terminal region is shown here.
Credit: Google DeepMind/EMBL-EBI (CC-BY-4.0)
A database containing the predicted structures of nearly every known protein on Earth has grown even larger and become more useful for understanding how the building blocks of life work together.
For the first time, the
AlphaFold protein-structure database
will include predictions of complexes of proteins — with the addition of 1.7 million ‘homodimers’ comprising two interacting strands of the same molecule.
The freely available database, maintained by the European Molecular Biology Laboratory’s European Bioinformatics Institute (EMBL-EBI) in Hinxton, UK, currently holds around 200 million predictions of individual protein structures, made using the AlphaFold2 AI tool, developed by London-based firm Google DeepMind.
Since its
release in 2021
, this repository has become a bedrock in discovery and a first port of call
for research projects
that try to understand life at the molecular level. But previous iterations of the database lacked predictions of how proteins form complexes, which can be indispensable for their function. For instance, HIV-1 protease — a viral protein that is a key drug target — works only when two copies of the same protein form a working enzyme.
AlphaFold is five years old — these charts show how it revolutionized science
Such proteins were already included in the database as individual ‘monomers’ but their entries tell only part of their story. “We thought, ‘can we bring the AlphaFold database to the next level, where we can include a lot of complex predictions across the tree of life?’” says Martin Steinegger, a computational biologist at Seoul National University in South Korea, who was part of the effort.
Complex interactions
To make predictions for even small complexes of two proteins was a crucial challenge, says Steinegger. “It is quite a different beast than monomer predictions.” Protein-complex predictions are exceedingly intensive computationally, so a consortium — including Steinegger’s lab, EMBL-EBI, Google DeepMind and chipmaker NVIDIA in Santa Clara, California — was formed to take on the challenge.
The consortium focused on protein complexes from 20 of the most studied species, including humans, mice, yeast and bacteria that cause disease in humans, such as
Mycobacterium tuberculosis
.
The huge protein database that spawned AlphaFold and biology’s AI revolution
Enjoying our latest content?
Log in or create an account to continue
Access the most recent journalism from Nature's award-winning team
Explore the latest features & opinion covering groundbreaking research
Access through your institution
or
Sign in or create an account
Continue with Google
Continue with ORCiD
doi: https://doi.org/10.1038/d41586-026-00787-3
References
Odai, R.
et al.
Sci. Adv.
11
, eadz8560 (2025).
Article
PubMed
Google Scholar
Download references
Reprints and permissions
Related Articles
The huge protein database that spawned AlphaFold and biology’s AI revolution
AlphaFold is running out of data — so drug firms are building their own version
Chemistry Nobel goes to developers of AlphaFold AI that predicts protein structures
AlphaFold touted as next big thing for drug discovery — but is it?
What’s next for AlphaFold and the AI protein-folding revolution
AlphaFold’s new rival? Meta AI predicts shape of 600 million proteins
Alphafold 3.0: the AI protein predictor gets an upgrade
AlphaFold is running out of data — so drug firms are building their own version
Beyond AlphaFold: how AI is decoding the grammar of the genome
Subjects
Machine learning
Drug discovery
Databases
Latest on:
Machine learning
Drug discovery
Databases
Rethinking AI’s role in survey research: from threat to collaboration
Correspondence
17 MAR 26
AI is programmed to hijack human empathy — we must resist that
World View
17 MAR 26
AI and the PhD student: friend or foe?
Technology Feature
16 MAR 26
Snapshots of the dynamic basis of NTSR1 G protein subtype promiscuity
Article
11 MAR 26
The age of animal experiments is waning. Where will science go next?
News Feature
25 FEB 26
‘An AlphaFold 4’ — scientists marvel at DeepMind drug spin-off’s exclusive new AI
News
19 FEB 26
Microsoft team creates ‘revolutionary’ data-storage system that lasts for millennia
News
18 FEB 26
An expanded registry of candidate cis-regulatory elements
Article
07 JAN 26
This science sleuth revealed a retraction crisis at Indian universities
News Feature
08 DEC 25
Jobs
PhD Candidate (m/f/d)
Profile the cellular components of tissue microenvironments (TME) in healthy and inflamed sites.
Dortmund, Nordrhein-Westfalen (DE)
Leibniz-Institut für Analytische Wissenschaften – ISAS – e.V.
Postdoctoral Positions in ‘Functional Microbiomics and Stem Cells’
Postdoctoral positions exploring microbiota–stem cell intera

---

## Expert Analysis

### Merits
N/A

### Areas for Consideration
- Complex interactions To make predictions for even small complexes of two proteins was a crucial challenge, says Steinegger. “It is quite a different beast than monomer predictions.” Protein-complex predictions are exceedingly intensive computationally, so a consortium — including Steinegger’s lab, EMBL-EBI, Google DeepMind and chipmaker NVIDIA in Santa Clara, California — was formed to take on the challenge.
- Meta AI predicts shape of 600 million proteins Alphafold 3.0: the AI protein predictor gets an upgrade AlphaFold is running out of data — so drug firms are building their own version Beyond AlphaFold: how AI is decoding the grammar of the genome Subjects Machine learning Drug discovery Databases Latest on: Machine learning Drug discovery Databases Rethinking AI’s role in survey research: from threat to collaboration Correspondence 17 MAR 26 AI is programmed to hijack human empathy — we must resist that World View 17 MAR 26 AI and the PhD student: friend or foe?

### Implications
- For the first time, the AlphaFold protein-structure database will include predictions of complexes of proteins — with the addition of 1.7 million ‘homodimers’ comprising two interacting strands of the same molecule.
- Where will science go next?

### Expert Commentary
This article covers alphafold, protein, database topics. Areas of concern are also raised. Readability: Flesch-Kincaid grade 0.0. Word count: 882.