Almonds as a Case Study for AI-Driven Compound Discovery

At the 2025 IFT FIRST Annual Event and Expo’s Hot Topic Studio session How AI and Metabolomics Are Discovering New Almond Bioactives, Dr. Josette Lewis, chief scientific officer at the Almond Board of California, spoke with Dr. Clay Bloszies, associate director of metabolomics at Brightseed. Their conversation highlighted how combining metabolomics with AI can accelerate compound discovery and support new scientific directions. Almonds served as a concrete example of how these tools can be applied across the food industry.

Lewis: To start, what does Brightseed do, and how does metabolomics fit into that work?
Bloszies: Brightseed focuses on uncovering health benefits in foods, plants, and microbes. Metabolomics lets us measure the full chemical profile of an ingredient rather than targeting a few known nutrients. From a single sample, we can detect hundreds to thousands of compounds. We then apply layers of AI trained on gold-standard public data across 23 health areas. That helps predict whether individual compounds may be bioactive and which systems they might influence. After that, we validate those predictions through in vitro or clinical work.

Lewis: Almonds have been the subject of clinical research for decades. What made them a useful case study?
Bloszies: Almonds are well researched but still revealed new findings when examined at a molecular level. We analyzed four California almond varieties and measured more than 500 compounds—about 200 lipids and 300 other metabolites. Roughly 400 had not been previously reported in almonds. We also identified 17 bioactive compounds across areas such as metabolism, immune function, and brain health, including six compounds known in the literature but never associated with almonds. That points to health areas—like oral health and joint health—that haven’t traditionally been explored in almond research.

Lewis: What do these findings show about the value of metabolomics for food ingredients more broadly?
Bloszies: They show how much remains undiscovered, even in familiar foods. The almond work demonstrates how metabolomics can reveal compounds and pathways that targeted analysis would miss. It also shows how a broader look at lipid species or polyphenols can expand understanding beyond established nutrient categories.

Lewis (commentary): We have 30 years of clinical and nutrition research showing almonds support cardiometabolic health—from weight maintenance to prediabetes markers. What we haven’t had before is molecular detail about the compounds that may contribute to those outcomes. This approach helps us understand mechanisms and identify new areas worth studying. It also shows how molecular insights can help guide future research directions for any ingredient, not just almonds.

Lewis: Where does AI fit into this process?
Bloszies: AI accelerates discovery by narrowing the funnel. Instead of testing thousands of compounds blindly, AI helps identify which ones are likely to be bioactive and what assays to run next. Its value depends on high-quality data—both in training and in how it’s interpreted. We see AI as a tool that helps prioritize research, not replace it.

Lewis: What makes the predictions more reliable over time?
Bloszies: We’ve profiled more than 5,000 plant species and built a database of more than 7 million compounds. That breadth helps refine and validate the models. The more we profile, the better the predictions become, and the more targeted our validation work can be.

Lewis: How might food companies use the outcomes of this kind of work?
Bloszies: There are several applications. Companies can use the insights to decide which health areas might merit funding for further research, shortening the path from discovery to clinical validation. They can use molecular detail to guide formulation or processing decisions—whether to preserve certain compounds, enhance them, or explore new combinations. And for companies pursuing future health claims, this work can help clarify what needs to be validated clinically.

Lewis: Did anything stand out in the almond work that could apply more broadly?
Bloszies: The range of nonlipid compounds with potential bioactivity was notable. We would expect certain polyphenols in berries, for example, but seeing substantial activity in almonds broadened our understanding. It reinforces that many foods may contain meaningful compounds that haven’t yet been identified.

Audience question: How could this research be used by the Almond Board of California to further consumer interest?
Lewis: As an industry, we began funding almond nutrition research 30 years ago, when almonds were not widely viewed as a healthy food. Today, in major markets around the world, consumers perceive almonds as nutritious. Our work now is to maintain relevance as health priorities evolve. These findings point to areas—such as joint health or oral health—that we can explore further. Molecular insights also help guide research that supports on-pack communication, whether for calorie content, prebiotic effects, or other emerging benefits. Ultimately, we serve as a resource for food companies, and research like this helps us provide evidence they can use in product development and marketing.ft

This article is based on a live session at IFT FIRST. Responses have been edited for length and clarity.