About

Summary of EVA TX Presentation at the International Society for Vaccines

October 11, 2026 - See Link for the Powerpoint as a PDF

 

The premise:

 

Computational tools can get you from a pathogen genome to a vaccine candidate in weeks rather than years. But prediction alone doesn't make a vaccine. What makes the design work is the human intelligence layered on top: three decades of judgment about which antigens are worth pursuing, which predicted epitopes are real, and which ones will quietly undermine the response you're trying to build. AI plus HI — that combination is what EVA Therapeutics is offering to vaccine developers in low- and middle-income countries.

 

Why this matters for LMIC vaccine development

 

EVA Therapeutics (EVA Tx) was established in April 2026 with a specific mission: to design and advance vaccines and immunotherapies targeting diseases that disproportionately affect LMIC populations — and to extend global access to personalized cancer vaccines.The model is deliberately structured so that design capacity sits alongside LMIC manufacturing and clinical capacity rather than replacing it. EVA Tx holds licenses to well-established vaccine design toolkits and technology (iVAX/ISPRI, Tregitope, Ancer) following the 2026 sale of EpiVax to SV Health, and works with a growing network of partners: the Center for Research on Global Health in Mali for clinical trials, Afrigen in South Africa for manufacturing, and development partners including CEPI and IVI, with additional collaborations in Brazil and Mexico.

 

The proposal De Groot put to the ISV audience was straightforward: EVA Tx wants to serve as the vaccine design partner to LMIC scientists — so that groups with manufacturing and trial capability don't have to build a computational immunology group from scratch to get a candidate worth making.

 

The toolkit

 

The iVAX toolkit was developed in 2009 under NIH U19 funding and is focused specifically on T cell epitope-driven vaccines (TEDV). Its components — EpiMatrix, ClustiMer, EpiAssembler, JanusMatrix, Conservatrix, OptiMatrix, VaccineCAD — do two distinct jobs: they accelerate design, and they optimize antigen sequences to improve the immune response. ISPRI provides detailed vaccine antigen analysis and ranks (and optimizes) antigens for vaccine design. 

 

Alongside these toolkits sits the Ancer personalized cancer vaccine platform, established in 2017 and exclusively licensed to EVA Tx globally. Ancer takes a patient's mutanome and genome and returns an HLA-matched multi-epitope vaccine — class I (CTL) and class II (T helper) epitopes, deliverable as peptides or mRNA, with tolerogenic epitopes stripped out by JanusMatrix. It is already in use in two private cancer vaccine clinics and EVA Tx is actively seeking more clinical partners.

 

The core ideas, briefly

 

Antigenicity is a function of T cell epitope content. Total predicted epitope content gives you a protein's immunogenic potential, and the EpiMatrix Immunogenicity Scale normalizes those scores so candidate antigens can be ranked against each other — and against familiar benchmarks like hepatitis B surface antigen (highly immunogenic), influenza HA (moderate), and the malaria RTS,S CSP antigen (low). Applied to Bundibugyo virus GP, the scale tells you immediately where your candidate sits in that range. (See PPT for more details). 

 

Not all predicted epitopes are the ones you want. This is where human intelligence does the heavy lifting. Pathogens and tumors engage in immune camouflage — carrying sequences that resemble human Treg epitopes, which recruit regulatory rather than effector responses. JanusMatrix evaluates cross-conservation against the human proteome to sort epitopes into three buckets: remove the ones extensively cross-reactive with self (probable Treg), keep the unique sequences (probable T effector), and deliberately include those cross-reactive with other circulating strains (T memory — protective against the next infection). An epitope prediction engine (read, "AI") without this step will confidently hand you sequences that suppress the response you need. That may be the issue with many vaccines that failed in the clinic. TBH. 

 

Conservation across strains buys breadth cheaply. Conservatrix identifies conserved 9-mers across many strains; EpiAssembler extends and overlaps those clusters into immunogenic consensus sequences. The result is protection against more strains using fewer epitopes — the published work (see our work in HIV vaccine design) shows that immunogenic consensus sequence epitopes are more cross-protective than strain-matched ones.

 

Then you build it. VaccineCAD computationally optimizes the order of epitopes in a "string of beads" construct, eliminating junctional artifacts, and the insert can be delivered as mRNA, as peptides, or on outer membrane vesicles — formats that partners like Afrigen and Intravacc can actually manufacture. 

 

Does it work? The published record

 

The approach has a long preclinical and translational track record:

 

•VennVax, a T cell epitope-based vaccinia/variola vaccine, gave 100% survival against a 10x LD50 intranasal challenge in HLA-DR3 transgenic mice, versus 19% in placebo, with strong Th1 cytokine responses across 50 epitopes. Of the predicted epitopes, 88% were confirmed in vitro using PBMC from Dryvax vaccinees. VennVax is now being developed as an Mpox vaccine at Afrigen. (Vaccine 2009;27:6471-9 and 2011;29:501-11)

 

•H. pylori: 100% cure in an infected mouse model. (https://www.sciencedirect.com/science/article/abs/pii/S0264410X11000119?via%3Dihub)

 

•Tularemia: protective against live challenge in a murine model.•Universal T cell epitope-driven influenza vaccine (UTIV) with Dan Hoft at Saint Louis University: immunogenic and protective using broadly conserved epitopes; preclinical and progressing toward the clinic.

 

•Multi-epitope COVID-19 vaccine: immunogenic in exposed and unexposed humans and in HLA transgenic mice.

​

•Lassa Fever: Lassa GP epitopes were validated in a demonstration project with Mark Poznansky at MGH/VIC, funded by DARPA.

 

On the immune engineering side, the H7N9 avian influenza work is the clearest demonstration of the principle. H7 HA scored unusually low on the immunogenicity scale; a predicted Treg epitope was identified and confirmed, then removed and replaced with T effector memory epitopes. The re-engineered HA produced a five-fold increase in antibody titer, remained stable on both Anhui/2013 and Guangdong/2016 backgrounds, and the unadjuvanted optimized construct outperformed wild type in challenge. (Wada et al., Sci Rep 2017;7(1):1283)

 

Underpinning all of it: class II HLA binding predictions show greater than 90% correlation with in vitro data, and prospective human validation studies span malaria, SARS-CoV-2, Q fever, mesothelioma, HCC, H. pylori, HFMD, influenza, smallpox and tularemia — for both CD4 and CD8 epitopes, plus a separate body of Treg epitope validation work.

 

Current and recent programs

 

Recent design work using tools now licensed by EVA Tx spans a genome-derived malaria vaccine (USAID, with Leidos/Oxford), a pan-coronavirus vaccine with OMV delivery (CEPI), the universal influenza A vaccine (SLU/NIAID), the Mpox T cell epitope string (CEPI), a patch-delivered COVID vaccine, and compassionate-use personalized cancer vaccines. Which vaccine will get to clinic first? We have to choose and so we did:

 

The active EVA Tx pipeline includes Mpox with Afrigen, Bundibugyo immune engineering with CRS Globale in Mali, a universal influenza vaccine with SLU (seeking partners), and patch delivery of tolerogenic peptides for allergy.The Lassa and Mpox collaborations illustrate how the division of labor is meant to work across the whole chain: design and project management with Jeff Safrit and Annie De Groot in the US; epitope validation and Phase I trial site with Ousmane Koita at CRS Globale in Mali; candidate production, manufacturing and pre-regulatory discussions with Caryn Fenner and Rajen Nadoo at Afrigen in South Africa; and immunogenicity and challenge studies with Wendy Burgers at the University of Cape Town.

 

The invitation

 

The bottleneck for neglected-pathogen vaccines has rarely been the absence of a genome. It's the long, tedious path from genome to a candidate worth manufacturing — and the fact that the computational capacity to shorten that path has historically sat far from the places where these diseases are endemic.EVA Tx is offering to close that gap as a design partner. If you're an LMIC vaccine developer with a pathogen, manufacturing capacity, or a trial site, the design step is available to you.

 

Annie S. De Groot, MD, President · Jeffrey T. Safrit, PhD, CEO