SEATTLE, WA, July 07, 2026 (GLOBE NEWSWIRE) — A-Alpha Bio, a synthetic biology and machine learning biotechnology company that measures and engineers protein-protein interactions, today launched its new web platform, Atlas. Atlas is a data ecosystem that enables researchers to prospectively generate abundant, high-quality binding data and access a repository of licensable, ML-ready protein interaction data. Together, these resources establish Atlas as the experimental data layer for AI-native protein engineering, offering the scale, diversity, and quality of data required to advance next-generation protein design models.
The new platform addresses a fundamental bottleneck in AI-enabled drug discovery: while antibody design models have progressed, their ability to reliably design high-affinity and developable antibodies remains constrained by limited access to large-scale experimental binding data. Antibodies and antigens exhibit extraordinary sequence and structural diversity, yet the affinity and structural datasets needed to advance modern AI models remain scarce.
“The Protein Data Bank (PDB) has been an invaluable resource over the past fifty years, but there are still fewer than 10,000 antibody-antigen structures against just 2,000 unique targets, and fewer than 800 that are paired with quantitative affinity data,” said David Younger, PhD, Co-Founder and CEO of A-Alpha Bio. “Generalizable models for de novo design and antibody optimization will require orders of magnitude more data than is available publicly. Atlas was purpose-built to fill this gap, and we believe it will be the industry’s go-to resource for training and benchmarking AI protein engineering models.”
Atlas is powered by AlphaSeq, A-Alpha Bio’s high-throughput platform for generating quantitative protein binding data. Since data is generated using the same experimental platform with standardized controls and consistent assay conditions, datasets can be combined across experiments for model training and benchmarking. This feature of Atlas data overcomes the major quality limitation of other binding datasets, which originate from different experimental methods and research groups.
Atlas gives researchers two ways to access protein-protein interaction data: licensing datasets that already exist in A-Alpha Bio’s Atlas database or generating new datasets designed to achieve specific research goals.
Data licensing: A-Alpha Bio is continuously migrating more of its 450 million affinity measurements and over 7,000 lab-validated antibody–antigen pseudo-structures into the Atlas database. These datasets are organized into application-specific ‘Data Blocks’ that are available for immediate, non-exclusive licensing.
Custom data generation: When off-the-shelf data is not the right fit, researchers can prospectively generate new AlphaSeq data through two options:
- Custom Data Blocks: datasets generated on-demand from customer-provided sequences, which are ideal for benchmarking the performance of in silico design strategies and training models on task-specific datasets. Custom Data Blocks can be provided on a private basis to accommodate proprietary sequences or targets.
- The Atlas Consortium: A-Alpha Bio’s answer to how model builders and pharmaceutical companies will affordably access the vast affinity and structural data required for training generalizable protein AI models. Subscriber and A-Alpha-designed Data Blocks are aggregated and distributed quarterly to all subscribers, delivering large-scale, diverse data at a fraction of the cost of independent generation. A VHH-specific Consortium is currently enrolling, with an scFV-focused Consortium expected to launch in 2027.
“We see Atlas as the experimental data layer for AI-native protein engineering,” said Randolph Lopez, PhD, Co-Founder & CTO of A-Alpha Bio. “The next wave of breakthroughs will depend on the tight integration between wet-lab and dry-lab workflows, where large-scale empirical data continuously pushes forward model development. By leveraging our unique ability to generate high-quality affinity and structural data at scale, we can help the industry accelerate towards realizing the promise of AI protein engineering.”
Explore Atlas at atlas.aalphabio.com to find the catalog of licensable Data Blocks, design a Custom Data Block, or join the Atlas Consortium.
About A-Alpha Bio
A-Alpha Bio is a biotechnology platform company that measures, predicts, and engineers protein-protein interactions at scale. Its experimental platform, AlphaSeq, enables the rapid and quantitative measurement of millions of protein-protein binding affinities simultaneously. Its AI platform, AlphaBind, is trained on the world’s largest protein-protein interaction database and predicts binding strength from sequence. A-Alpha Bio works with pharmaceutical and biotechnology companies through both strategic partnerships and streamlined services, leveraging AlphaSeq and AlphaBind to engineer next-generation dual-specific and molecular glue therapeutics, discover and optimize differentiated biologics, and generate protein binding and structural data at scale to power AI models. A-Alpha Bio is based in Seattle and was founded in 2017 at the University of Washington’s Institute for Protein Design. To learn more, please visit: https://www.aalphabio.com/
Media Contact:
Drew Duglan
Director, Scientific Communications, A-Alpha BIo
dduglan@aalphabio.com
858-247-9110
