AlphaFold 2 reached more than three million researchers across 190 countries by November 2025, marking unprecedented adoption for an AI scientific tool. The protein structure prediction system developed by Google DeepMind earned its creators the 2024 Nobel Prize in Chemistry after transforming structural biology research. The database now contains over 214 million predicted protein structures, expanding from 300,000 entries at its July 2021 launch.
AlphaFold 2 Key Statistics
- AlphaFold 2 serves more than 3 million researchers globally as of November 2025, with over 1 million users from low and middle-income countries.
- The database contains 214 million protein structure predictions spanning 23 terabytes of data as of 2024.
- AlphaFold 2 achieved a median GDT_TS score of 92.4 at CASP14, reaching experimental-level accuracy for 63% of evaluated protein domains.
- Research publications utilizing AlphaFold grew at an annual rate of 180.13% from 2019 to 2024, with the primary paper cited approximately 43,000 times.
- The system doubled the ratio of druggable protein pockets from 19.8% to 41.8%, significantly expanding pharmaceutical research targets.
AlphaFold 2 Database Growth and Global Adoption
The AlphaFold Protein Structure Database expanded 500-fold between July 2021 and 2024. The initial release contained approximately 300,000 predicted structures covering proteins primarily from model organisms.
Google DeepMind and EMBL-EBI scaled the database to 214 million entries by 2024, representing nearly all catalogued proteins known to science. The dataset requires 2.5 days for complete download using a 1 Gbps connection and remains freely accessible under a CC BY 4.0 license.
The database recorded 23,000 archive downloads and attracted 1.6 million unique visitors by January 2024. Over 30% of AlphaFold 2 usage focuses on disease-related research according to 2025 data from Google DeepMind.
| Metric | Value | Date |
|---|---|---|
| Total Protein Structures | 214 million+ | 2024 |
| Database Size | 23 TiB | 2024 |
| Active User Countries | 190+ | November 2025 |
| Archive Downloads | 23,000+ | January 2024 |
AlphaFold 2 Prediction Accuracy and Performance
AlphaFold 2 achieved a median Global Distance Test Total Score of 92.4 at the CASP14 competition in 2020. The system reached experimental-level accuracy for 58 of 92 evaluated protein domains with GDT_TS scores exceeding 90.
The z-score differential demonstrated substantial performance superiority. AlphaFold 2 recorded 244.0 compared to 90.8 for the second-best competitor, representing three times greater accuracy than traditional computational methods.
Confidence Score Distribution
The predicted Local Distance Difference Test score provides per-residue confidence metrics. Predictions with pLDDT scores of 90 or higher show a median root-mean-square deviation of only 0.6 Ångstroms when compared against experimental structures.
| pLDDT Range | Confidence Level | Median RMSD |
|---|---|---|
| ≥90 | Very High | 0.6 Ångstroms |
| 70-90 | Confident | 1.0-1.5 Ångstroms |
| 50-70 | Low | 2.0-3.0 Ångstroms |
| <50 | Very Low | 3.5 Ångstroms |
AlphaFold 2 Impact on Drug Discovery
AlphaFold 2 models expanded the druggable proteome by revealing accessible drug pockets across previously intractable protein targets. Research from canSAR demonstrated an increase from 19.8% to 41.8% in druggable protein ratios.
Virtual screening studies for TAAR1 receptors achieved a 60% hit rate using AlphaFold-based models. Traditional homology modeling approaches reached approximately 30% hit rates under comparable conditions. Identified compounds demonstrated potencies ranging from 0.03 to 12 micromolar.
Industry analysts estimate AlphaFold could reduce typical drug development timelines from ten years to seven years. The acceleration primarily affects the target identification and validation phases of pharmaceutical research.
AlphaFold 2 Human Proteome Coverage
AlphaFold 2 doubled the structural coverage of human proteins compared to experimentally determined structures in the Protein Data Bank. The system generated high-confidence predictions for 35.7% of residues in the human proteome.
Coverage reached 58% of residues with confident predictions above pLDDT 70. The database includes 1,290 human proteins containing substantial regions exceeding 200 residues with confident predictions and no suitable experimental template available.
AlphaFold 2 Research Applications and Nobel Recognition
Demis Hassabis and John Jumper received half of the 2024 Nobel Prize in Chemistry for developing AlphaFold 2. The Royal Swedish Academy of Sciences recognized artificial intelligence’s contribution to scientific discovery for the first time in the Chemistry category.
University of Oxford researchers utilized AlphaFold 2 to determine the complete structure of the Pfs48/45 malaria surface protein. The breakthrough enabled clinical trials for transmission-blocking vaccines to begin in 2023 after years of unsuccessful attempts using traditional experimental methods.
The Drugs for Neglected Diseases Initiative incorporated AlphaFold predictions for Chagas disease and leishmaniasis research. The organization maintains a portfolio of over 20 new chemical entities targeting conditions affecting developing countries.
| Application Area | Achievement |
|---|---|
| Malaria Research | Complete Pfs48/45 protein structure revealed |
| Enzyme Engineering | Plastic-degrading enzyme development |
| Neglected Diseases | 20+ new chemical entities in DNDi portfolio |
| Antibiotic Resistance | Enhanced understanding of resistance mechanisms |
FAQ
How many researchers use AlphaFold 2?
More than 3 million researchers from 190 countries use AlphaFold 2 as of November 2025, with over 1 million users based in low and middle-income countries.
How accurate is AlphaFold 2?
AlphaFold 2 achieved a median GDT_TS score of 92.4 at CASP14, reaching experimental-level accuracy for 63% of protein domains with median errors under 1 Ångstrom.
How many protein structures are in the AlphaFold database?
The AlphaFold database contains over 214 million predicted protein structures as of 2024, representing a 500-fold expansion from the initial 300,000 structures at launch.
Did AlphaFold win a Nobel Prize?
Yes, Demis Hassabis and John Jumper received the 2024 Nobel Prize in Chemistry for developing AlphaFold 2, marking the first AI contribution recognized in this category.
How does AlphaFold 2 impact drug discovery?
AlphaFold 2 doubled the ratio of druggable protein pockets from 19.8% to 41.8% and achieved 60% hit rates in virtual screening compared to 30% for traditional methods.
