Ai-accelerated discovery of natural sweeteners from over 1M natural compounds
Sensorygen uses AI algorithms to screen over 1 million natural compounds, identifying novel sweeteners with desirable commercial properties. The platform models human taste perception, predicts sweetness with high accuracy, and evaluates long-term safety profiles, enabling faster, cost-effective discovery of natural sugar alternatives for food and beverage applications.
Overview
Sensorygen offers an AI-driven platform for discovering natural sweeteners by screening a vast chemical space of over one million natural compounds. The technology models human taste and olfactory perception to identify novel sweeteners tuned for consumer use, including long-term safety profiling. This approach enables companies in the food, beverage, and flavor industries to accelerate the identification of natural sugar alternatives while reducing reliance on traditional trial-and-error laboratory methods. Candidates identified through the platform are evaluated against commercial criteria such as odor profile and activity on sweet, bitter, and umami receptors.
Technical specifications
- AI-based screening of over 1 million natural compounds to identify candidates with high commercial potential
- Sweetness prediction algorithms with a correlation of approximately 0.90 between estimated and actual sweetness, and a mean absolute error of 0.4 in predicting log sweetness values
- Multi-receptor profiling to evaluate predicted activity on sweet, bitter, and umami receptors
- Safety and regulatory filtering including odor/taste profiling, toxicity estimation, and approval path assessment
- Preliminary screening of approximately 270,000 natural compounds has already yielded patented sweetener candidates and positive allosteric modulators of sweet receptors
- Future validation pipeline includes human taste bud epithelial cell (htBEC) assays via Eurofins contract research services and human sensorial testing panels, with optional behavioral assays in mice
Technology readiness level
The platform has demonstrated validated predictive accuracy for sweetness, with algorithms achieving strong correlation to actual sensory data. Preliminary compounds have been identified from an initial screening of approximately 270,000 natural compounds, and patents have been filed covering sweetener candidates and positive allosteric modulators. The next phase involves evaluating 100 predicted sweeteners using cell-based assays and human sensorial testing over an estimated 12-month period, with the goal of delivering a validated set of prospective natural sweeteners ready for commercial development.
About Sensorygen
Sensorygen is a life science company that utilizes computational biology and artificial intelligence to discover natural compounds with improved efficacy, safety, and sensorial properties. Founded in 2015 by Dr. Anandasankar Ray, the company developed AI-driven platforms such as SensoryAI and ToxAI, which enable the rapid screening of millions of compounds to identify natural replacements for toxic or harsh ingredients. These technologies are applied across diverse sectors, including consumer products, agriculture, flavors, fragrances, and toxicology, allowing for more efficient, digital-first ingredient discovery that reduces reliance on traditional, trial-and-error laboratory methods and animal testing.
This technology provides significant value to partners and customers by helping them future-proof ingredient portfolios, reduce development costs, and accelerate the time-to-market for safer products. By predicting how chemicals interact with sensory receptors, Sensorygen helps brands optimize for consumer-preferred attributes like scent and taste while ensuring regulatory compliance and environmental safety. The company has secured $1.5 million in equity funding and maintains an intellectual property portfolio licensed from the University of California, Riverside, supporting its efforts to deliver high-performance, non-toxic alternatives for everyday applications.
delivered to your inbox.