Computational Design of Smart Antibodies

Yanay Ofran, PhD, Chief Executive Officer, Biolojic Design

Yanay Ofran, PhD, provided a compelling presentation regarding artificial intelligence (AI) platforms and their potential to customize and enhance the natural behavior of antibodies. The research aim of Biolojic Design is to explore the creation of smart molecules that respond to diverse environments intellectually by design vs. a random antagonistic behavioral response.

Programming Human Antibodies: Dual MOA Applications

The research objectives and direction that Ofran presented focused on gaining control of certain antibody functions by predetermining the antagonistic responses and eliminating happenstance-type reactions. The vehicle for creating these types of circumstance-independent antibodies is based on two arms or application models.Eliciting the antibody behavior in such a way allows for a predictive, controlled responses relative to:

Binding to a specific surface of the target (an epitope specific antibody)

Allowing for stabilization of a specific confirmation.

Multi-specificity.

Then can use antibodies to program biology.

Designing Antibodies with Enhanced Functional Capabilities

The first computational designed Ab is in the clinic undergoing Phase 1 trials collaborating with various biotech and pharmaceutical entities. The design process also featured a proprietary methodology that involved the collection of millions of antibody measurements performed over a decade. Additional proprietary gains include platform IP specific, the composition-of-matter protection for products. The scope of novel capabilities provided by these designed antibodies include:

Agonistic behavior by design

Conditional function(if/then)

Multi-bodies or (gates)

Sensors and Actuators

Example of AU:007 (anti IL2 Antibody): Mimicking the Nature of the Immune System and How It Produces Antibodies

MethodsIL2 treatments in clinic consisted of administering the patient modified IL2 which was successful and created a negative feedback loop.ChallengesThe challenging engineering wise is hard, as you are trying to achieve a surgical type binding and high concentrations lead to toxicities expressed on lung cells, blood vessel leakage, and pulmonary edema.ResultsCMC

Completed CMC in 7 months

Yield. >5g/L

Extremely stble, no aggregation, no PTM

Highly specific

Toxicity Profile

No-tissue cross activity reported.

No toxicity in NHP repeat dose.

Low immunogenicity in primates.

Clinical Findings

Phase 1 study in progress and the first in-human trial is upcoming.

Understanding BD9s Relativity to Similar Therapeutics Already in Clinic: A Comparative Conclusion

After analyzing the BD9 therapeutic model compared to a clinically approved antibody. It was found to be slightly better than similar approved antibody models. Though better or worse wasn’t the desired takeaway, but merely an attempt to validate that the BD9 therapeutic can deliver similar and promising results as a poly-specific antibody type therapeutic.

Summary: Using AI to Help Discovery

AI can give antibodies new capabilities that traditional methods cannot do. They can measure a dominant cytokine and allow for an antibody to determine its function based upon specific micro-environments circumstance. Engineering next generation applications might include further enhancements on developing CDRs that are not as rigid, using new residues, and improved KD values.