Tevogen.AI Boosts PredicTcell Accuracy to 92% Recall and 48% Precision
Tevogen.AI's PredicTcell beta increased recall from 87% to 92% and precision from 40% to 48% using 1.8 million training points, 20x more data than its initial model. The platform draws on a 655 million-sequence database (16 billion data points) and uses three AI agents for continuous learning for pharma partnerships.
1. PredicTcell Beta Performance
Tevogen.AI’s PredicTcell beta model achieved a recall rate increase from 87% to 92% and precision improvement from 40% to 48% by leveraging 1.8 million training data points, yielding higher true positives and fewer missed targets compared to its initial version.
2. Proprietary Peptide Database
The platform now incorporates a proprietary peptide database of 655 million sequences derived from 24 million proteins, totaling 16 billion data points, with continuous enrichment through analysis of 37 million scientific publications to enhance target prediction across disease areas.
3. Continuous AI Learning Agents
Three production AI agents continuously evaluate 14 active peptide candidates, monitor newly published scientific literature, and integrate wet lab validation results back into the AI system, creating a feedback loop that strengthens predictive accuracy over time.
4. Partnership Strategy and Outlook
With improved predictive metrics and a robust data infrastructure, Tevogen plans to engage pharmaceutical partners to advance select peptide candidates into production and development, aiming to reduce development costs, accelerate time to market, and increase clinical success rates.