Announcing a major upgrade to its AI platform for drug-combination discovery, Algorae Pharmaceuticals Ltd launched AlgoraeOS Version 2, which was developed in partnership with researchers from the UNSW Biomedical AI Laboratory, the UNSW AI Institute, and with support from CSIRO Data61.

This new version is positioned as a significant milestone, setting a new benchmark for identifying effective drug combinations in complex diseases and serving as the cornerstone of the company’s R&D strategy.

AlgoraeOS v2 has been trained at scale on over 5.5 million unique inhibition records and engineered to model the full dose-response surface across four standard synergy measures—Bliss, Loewe, HSA, and ZIP—thereby replacing less informative single-metric, dose-averaged summaries with a dose-resolved assessment of efficacy. Crucially, the platform demonstrates benchmark leadership, having outperformed representative state-of-the-art models, including Google DeepMind's TxGemma-27B-Predict and Tx-LLM (M), consistently achieving lower error and stronger correlations across synergy formalisms in published benchmarks like NCI-ALMANAC. The platform's superiority stems from its use of uncertainty-aware deep learning, which provides confidence-weighted predictions for every output, enabling risk-aware experimental design by reporting both the predicted outcome and its certainty. This capability makes AlgoraeOS v2 a more reliable, decision-grade tool that can accurately predict synergy for previously unseen drug combinations. Algorae plans to immediately deploy this cutting-edge platform across its preclinical pipeline to guide hit prioritization, dose selection, and study design for high-potential combination therapies, with the first in-silico fixed-dose combination (FDC) predictions generated by AlgoraeOS v2 expected in Q4 2025.