Overcoming the differences in post-translational protein modification mechanisms between prokaryotes and eukaryotes, such as glycosylation and disulfide bond formation, have been a technological…

The Department of Agriculture Animal and Plant Health Inspection Service awarded a $1.998 million cooperative agreement to Georgia TECH Applied Research Corp on February 2, 2026, under the Plant and Animal Disease, Pest Control, and Animal Care program (CFDA 10.025).

Georgia TECH Applied Research Corp will develop engineered probiotic bacteria-based influenza vaccines using artificial intelligence and protein structure prediction methods. The work addresses post-translational protein modification barriers preventing accurate antigenic expression of viral epitopes in bacterial systems, particularly for influenza hemagglutinin (HA) spike protein, which requires proper trimerization, glycosylation, and disulfide bond formation. The recipient will apply multi-agent frameworks based on large language models and AlphaFold protein modeling to mine historical HA sequence and antigenic structure data, perform in silico evaluation and mutation modeling to identify amino acid substitutions compensating for differential bacterial protein expression, and synthesize candidate sequences as degenerate oligo pools for cloning into expression plasmids. Successful candidates will be tested in parallel using a high-throughput bacterial outer membrane protein plasmid-based expression and screening system capable of evaluating millions of sequences simultaneously through down-selection with known HA-inhibiting antibodies. The objective is to develop low-cost, shelf-stable animal vaccines administrable through feed that elicit protective antibodies against influenza, including highly pathogenic avian influenza (H5N1), with potential for multivalent vaccine formulations combining multiple engineered probiotic bacterial strains.

Performance occurs in Georgia with a period extending from February 2, 2026, through February 1, 2029.

Overcoming the differences in post-translational protein modification mechanisms between prokaryotes and eukaryotes, such as glycosylation and disulfide bond formation, have been a technological obstacle preventing the structurally accurate 'antigenic' expression of viral epitopes in bacteria. This problem is exacerbated in the case of influenza's hemagglutinin (HA) spike protein as its correct antigenic structure is influenced by trimerization, glycosylation, and DI-sulfide bond formation. If this system structural incompatibility can be overcome, probiotic bacteria could become a low-cost, shelf-stable animal vaccine that could be administered through animal feed and elicit protective antibodies against influenza in general and highly pathogenic avian influenza (H5N1) in particular. Indeed, a multivalent vaccine made up of mixes of engineered probiotic bacteria is possible. During this project, we will develop and combine methods from artificial intelligence and biomolecular modeling, such as multi-agent framework based on large language models (llms) and the nobel-prizewinning alphafold program, to mine historical HA sequence and antigenic structure information. This will enable the in silico evaluation and modeling of potential mutations or amino acid substitutions needed to compensate for the differential bacteria protein expression system to reproduce key antigenic structures. These candidate sequences may be in the form of a monomer, trimers, antigen-scaffold fusions, and fusions to orthogonal bacterial membrane protein domains. Successful in silico-developed HA sequences will be synthesized as degenerate oligo pools, cloned into expression plasmids, and tested in parallel in vitro with a developed high throughput bacterial outer membrane protein plasmid-based expression and screening system. This system can screen millions of candidate sequences simultaneously through down selection with known HA-inhibiting antibodies. Protein sequences that successfully present antigen in the plasmid expression system will then be inserted directly into the bacterial genome eliminating the need for plasmids, which carry unwanted artificial antibiotic markers and other unnatural genetic sequences. The ultimate objective of this project is to develop a poultry vaccine production system, that is low cost with a long shelf life at ambient temperatures and that can be rapidly manufactured and distributed as a probiotic feed supplement. This system would be highly multivalent and rapidly reconfigurable for new or emerging viral threats.