Designed to help biopharmaceutical teams focus on the most promising drug candidates, EVA generates clinical predictions with up to twice the performance of current industry standards.
Scienta Lab, leader in precision immunology, today announced the launch of EVA, the first multimodal AI model purpose-built to support drug discovery and development in immunology and inflammation. EVA addresses a central challenge faced by the biopharmaceutical industry: translating biological signals generated in preclinical research into robust decisions that increase the probability of clinical success.
A model designed for translational R&D
EVA is an artificial intelligence model specifically designed to tackle translational challenges in immunology and inflammation, supporting decision-making across the entire drug development lifecycle.
At the earliest stages of R&D, EVA supports the prioritisation of therapeutic targets by estimating their efficacy based on a simple description of the drug candidate’s mechanism of action. As programs progress, EVA assesses the robustness and translatability of preclinical biological signals, including those derived from murine models, by evaluating how molecular perturbations propagate across various therapeutic indications. At the clinical stage, the model supports the identification of patient subpopulations most likely to respond to treatment, contributing more efficient clinical trials.
Pre-trained at scale on data spanning multiple immuno-inflammatory diseases, EVA operates without the need for task-specific retraining, enabling fast and consistent deployment across the R&D pipeline.
“EVA has been designed to answer the concrete questions faced by immunology R&D teams: which targets to prioritize, which preclinical signals are truly translational, and which patients are most likely to respond to a drug candidate,” says Julien Duquesnem CTO and Co-Founder of Scienta Lab. “The goal is not to replace the experimentation, but to better guide decision-making at every stage of the development.”
Performances up to two times higher than current industry standards
From a technological standpoint, EVA is built on a multimodal foundation model architecture. The model learns patient-scale biological representations by integrating more than half a million human and murine transcriptomic, histological, and proteomic samples.
Benchmarked against state-of-the-art AI models and reference statistical approaches across 40 use cases covering the full drug development process, EVA demonstrates performance gains of up to twofold compared to current industry standards. These results enable biopharmaceutical companies to make earlier and more informed decisions, reduce late-stage failures, and ultimately accelerate patient access to effective therapies.
The scientific foundations of EVA and detailed benchmarking results are described in a preprint now available on arXiv. Scienta Lab has also released an open version of its transcriptomic model, comprising 60 million parameters, on the Hugging Face platform to support research in computational immunology. Large-scale deployments and customized applications continue to be offered through commercial partnerships.
The article EVA: Towards a universal model of the immune system, is available here.
About Scienta Lab
Founded in 2021 in Paris, Scienta Lab is a company specializing in the application of artificial intelligence to drug discovery and development in immunology and inflammation. The company developed EVA, an AI model designed to support biopharmaceutical stakeholders in decision-making across the R&D process. In 2025, Scienta Lab was selected for the EIC Accelerator program, the European Union’s flagship initiative for breakthrough technologies.
