Evaxion Biotech A/S is leveraging two artificial intelligence (AI) platforms to quickly identify targets for cancer immunotherapies and infectious diseases. The Copenhagen-based company's Pioneer platform identifies the tumor-specific mutations most likely to stimulate a strong response from a patient's own immune system to fight cancer, allowing development of truly individualized therapies within seven weeks. The Eden platform identifies novel bacterial vaccine antigens designed to produce robust immune responses to a range of bacterial substrains in a few hours.
The company received $3 million in December 2017 from Innovation Fund Denmark to quickly and cost-effectively develop a personalized cancer vaccine by combining gene technology and the Pioneer AI platform which integrates neural networks, big data and supercomputing. The initial dose of the company's first immunotherapy, EVX-01, was administered in April.
The biotech company is not shy about its goals. Evaxion "aspires to decode the complexity of the human immune system using machine learning so we can program the immune system to elicit the response we want," said Evaxion's co-founder and chief operating officer, Niels Møller. "On our way to achieving our aspiration, we will radically change the drug discovery process, with the aim of defining immunotherapies and significantly improving the lives of patients."
The Pioneer AI rapidly processes data from DNA sequencing of both tumor biopsies and patient blood samples to distinguish the somatic variants or neoepitopes found only in the tumor. Those are then further analyzed to identify the cancer-specific neopeptide sequences. The AI platform predicts how likely each peptide sequence is to be "processed and presented by the tumor cells and antigen presenting cells," explained company CEO Lars Wegner. "These predictions are used in a unique combination of features of treatment efficacy to rank each neoepitope."
The AI then ranks the sequences for likely immunogenicity or strength of T-cell response. The top ranked choices are selected for production, purified, and analyzed for good manufacturing practices before releasing the final drug product. Møller told Bioworld the process takes no more than seven weeks from the date of biopsy to delivery of a personalized therapy to the patient.
"Pioneer is designed for truly personalized immunotherapy. We have not yet produced the same neoepitope for two different patients," Møller said. "Evaxion has assessed the possibility to identify cancer neoepitopes shared across patients. However, only a very limited number of neoepitopes are shared between different patients."
The Eden AI takes aim at antigens that have evaded the immune system, what Wegner called the "Achilles heel" of pathogens. Evasion leaves "fingerprints" over time on the antigens. Eden identifies those tell-tale markers through machine learning, allowing it to identify potential novel protective vaccine proteins.
After receiving input of the DNA sequence of a specific pathogen, the Eden platform is trained to find antigens that elicit a strong protective immune response against bacterial infections, by translating amino acid sequences in the bacterial proteome into feature vectors using in silico feature predictors.
"One feature, for example, is disulfide bonds and major histocompatibility complex binding sites," Møller said. "Based on the training set, Eden has learned to differentiate between particular patterns in the feature vectors of proteins that drive a protective immune response from those that do not. The feature vectors of the proteins that share patterns with the protective proteins in the training set are identified as protective proteins by Eden."
Eden completes the full analysis of all proteins in the proteome within 48 hours, along with ranking of the potential protective antigens.
"Following identification, Evaxion synthesizes the top 20 ranking proteins and tests these in relevant animal challenge models," Wegner said. "Antigens that consistently show high levels of protection are selected for further development and characterization."
The company typically combines antigens with complementary functionalities when developing vaccine candidates, increasing their effectiveness. "Generally, the top-ranking proteins identified by Eden are very effective, providing close to 100 percent protection in different animal models."
Eden has identified and validated potential protein vaccine antigens across a wide range of pathogen types, including gram-negative, gram-positive, extracellular and intracellular bacteria as well as some parasites in preclinical animal models, Wegner noted.
Eden can be used to develop vaccines against multidrug-resistant pathogens and non-resistant bacterial strains as well as in emerging epidemics of new zoonotic pathogens. "The prophylactic nature of vaccines will preserve antibiotic use and in turn prevent further development of resistance to antibiotics," said Møller.
"Evaxion has infectious disease pipeline programs which target S. aureus, P. aeruginosa, non-typeable Haemophilus influenza and Moraxella catarrhalis," he added. "Evaxion has successfully validated Eden in the discovery of vaccines with bioterrorism and biowarfare potential as well as parasitic and neglected diseases. However, for the time being, we are not pursuing development programs in these areas."