The Ordaōs Design Engine is a multitask meta-learning model that leverages continuous learning loops and proprietary data sets to translate human-defined product criteria into machine-designed miniPRO™.

Starting with amino acids, the engine generates, appraises, and ranks billions of protein sequences, hundreds of thousands of protein structures and other drug-like properties to create miniPRO™.

miniPRO™ are evaluated to provide intelligent feedback on multiple design objectives including protein structure, binding specificity and affinity, solubility, stability, immunogenicity, and developability. The analysis feeds the engine within a reinforcement learning loop, to iteratively improve and deliver optimal miniPRO™ that meet the client’s molecular target product profile (mTPP).