Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

David Younger explains that training generalizable protein design models requires hundreds of thousands of structural affinity landscapes. Even using high-throughput tools that measure a million affinities at once, funding that volume alone is nearly impossible for one firm, whereas a five-member consortium yields 28 million data points per year.

Related Insights

The cost to generate the volume of protein affinity data from a single multi-week A-AlphaBio experiment using standard methods like surface plasmon resonance (SPR) would be an economically unfeasible $100-$500 million. This staggering cost difference illustrates the fundamental barrier that new high-throughput platforms are designed to overcome.

David Younger divides the data bottleneck into generalizable model training and proprietary candidate validation. Broad target and binder diversity can be generated collaboratively in pre-competitive pools, while validating clinical assets against specific proprietary drug targets remains entirely private and isolated outside the consortium.

Machine learning's application in multi-specific antibody design is hampered by a lack of public data. Companies must invest heavily in generating their own large-scale, proprietary datasets to train effective models, creating a significant barrier to entry and a competitive advantage.

To break the data bottleneck in AI protein engineering, companies now generate massive synthetic datasets. By creating novel "synthetic epitopes" and measuring their binding, they can produce thousands of validated positive and negative training examples in a single experiment, massively accelerating model development.

Unlike purely in-silico companies, Metaphor's platform starts with high-throughput wet lab experiments to generate massive datasets on receptor interactions in living systems. This real-world data is crucial for training their AI to design functionally active antibodies.

Younger says model makers evaluate on unique bespoke targets, so pharma buyers cannot tell which model is best. He is working with companies on a benchmark that standardizes target difficulty and characterizes tens of thousands of designs, hoping to get true apples-to-apples comparisons with tighter error bars.

Current AI for protein engineering relies on small public datasets like the PDB (~10,000 structures), causing models to "hallucinate" or default to known examples. This data bottleneck, orders of magnitude smaller than data used for LLMs, hinders the development of novel therapeutics.

Younger notes the PDB shows how proteins physically interact but not how strongly, and it has almost no examples of antibodies one or two mutations apart with measured affinities. Zero-shot optimization instead needs local affinity landscapes built from many single-point mutation variants across many targets.

David Younger says his company's model training showed that solving truly hard protein design problems will require much more data than the field expects. He also warns public binding and affinity data are often not interoperable, lack proper controls, and lack annotations, making them hard to reuse.

Younger says the Protein Data Bank today holds about 10,000 antibody-antigen structures and adds roughly 1,000 yearly. Their internal scaling work suggests good generalizable de novo design models may need hundreds of thousands to millions of structures, so waiting on PDB growth is far too slow.

Antibody ML models demand data scales exceeding single-company budgets | RiffOn