InVirtuoLabs sets a new standard in generative drug discovery

InVirtuoLabs, a biotech startup incubated at USI Startup Centre that merges deep learning with molecular science, has released a new foundational model, InVirtuoGEN, marking a breakthrough in AI for pharmaceutical research.

InVirtuoGEN, the new model of InVirtuoLabs, represents a significant step forward in the generative drug discovery field, outperforming other models, including the one of NVIDIA, on every benchmark. It can generate entirely new chemical compounds, enhance existing molecules, and optimise binding to target proteins. The work behind InVirtuoGEN will be presented at the NeurIPS AI for Drug Discovery Workshop, part of one of the world’s leading machine learning conferences, held in San Diego in December 2025.

InVirtuoGEN continuously refine structures, improving promising drug candidates step by step. This iterative approach results in more realistic, diverse compounds while drastically reducing both computational costs and development time. The model has already demonstrated impressive results in various benchmarks:

• De novo molecule generation: A 35% improvement in success rate for generating viable drug-like compounds, compared to previous state-of-the-art methods.
• Existing molecules optimisation: A 40% improvement in maintaining structural fidelity, alongside increased molecular diversity in fragment-constrained design tasks
• Protein binding optimisation: A 20–30% improvement in binding scores and faster convergence, consistently outperforming any competing system, including NVIDIA’s reference model.

InVirtuoGEN has already been implemented in the InVirtuoLabs’s internal drug discovery pipeline and is also being integrated into the InVirtuoPlatform to support partner pharmaceutical programmes. With this groundbreaking achievement, InVirtuoLabs is solidifying its position as a leader in AI-driven biotechnology, turning cutting-edge machine learning research into tangible solutions for patients and partners worldwide.