Marijuana Breeders Can Use AI To Design New Strains, Study Demonstrates

Marijuana breeders may be able to design new strains and speed up their growing cycles by utilizing artificial intelligence (AI), a new study suggests.

Researchers found that by feeding genetic markers, growth measurements, environmental data and chemical assays into AI models, breeders could simulate thousands of potential crosses and stimulate “speed breeding” through machine learning before ever planting a seed.

The authors argue this approach could cut traditional breeding cycles, which currently last between six to eight years, down to a fraction of that time, while also improving consistency—a perennial challenge that commercial cannabis growers grapple with.

“Machine learning allows for iterative simulations of breeding outcomes…while ensuring chemical consistency,” the authors concluded.

“AI-enabled cannabis breeding represents a paradigm shift in strain development, enabling precise control over cannabinoid and terpene profiles while reducing breeding cycle times and resource requirements.”

The paper also highlights the role of metabolomics, an emerging field that catalogs the vast array of chemicals produced by living organisms.

“AI systems correlate these datasets to predict how specific genetic combinations will influence chemical composition and growth traits, enabling precise selection of parental strains for crossbreeding,” they observed.

Techniques like genomic selection, regression analysis and deep learning are already used in major agricultural crops. Applied to cannabis, these tools look for patterns linking genetic variants to chemical traits such as the proportion of THC or the presence of rare cannabinoids like CBG.

One of the biggest challenges in cannabis cultivation is the complex interaction between genetics and environment. Light spectrum, humidity, nutrient availability and subtle temperature shifts can reshape a plant’s chemical output.

The study, which has not yet been published in a journal but was posted on the science site ResearchGate, describes how AI systems can incorporate these variables to predict performance in different growing environments—a tool that could be particularly valuable as the industry expands into diverse climatic regions.

Neural networks can track nonlinear interactions among dozens of genes that influence plant chemistry, in addition to incorporating environmental conditions.

“These innovations promise accelerated strain development, improved chemical consistency, and enhanced adaptability to diverse growing environments,” the authors, who are affiliated with the University of Saskatchewan and Renaissance Bioscience, noted.

Using instruments like gas chromatography mass spectrometers, researchers can employ AI to measure cannabinoids and terpenes across a plant’s life cycle. Combined with imaging tools that assess traits such as trichome density or stress responses, these data points give AI models the raw material to make increasingly accurate predictions.

“This capability allows breeders to design strains not only for chemical profiles but also for resilience and adaptability in diverse growing environments,” the researchers wrote.

Highlighting the importance of reproducibility, they noted that “the global cannabis industry demands high-quality, reproducible strains, creating the need for precision breeding technologies that reduce time-to-market while maximizing yield and potency.”

The writers cautioned that, for all its advantages, AI-enabled cannabis breeding faces challenges, including data quality constraints that affect the accuracy of genotypic and phenotypic predictions. They also note complications surrounding complex polygenic traits, ethical considerations and regulatory barriers, observing that “legal restrictions on cannabis research may limit data access.”

Recent research on cannabis genetics suggests that incentives in the legal marijuana market—such as the desire for plants to mature faster and produce more cannabinoids for extraction—may be leading to a decline in biodiversity of the plant worldwide, prompting a researcher from California State Polytechnic University Humboldt to describe the problem as “the bottlenecking of Cannabis genetics.”

And in 2022, the California Department of Cannabis Control funded a $20 million study to, as officials wrote, “identify and preserve the history, value and diversity of California legacy cannabis cultivars and the rich experience of its legacy cultivation community, and enable, enhance and guide the understanding and application of cannabis genetics to the greater body of research and science-based public policy development.”

Interest in cannabis genetics is not confined to state level governments. In 2018, a powerful U.S. Senate committee directed agriculture authorities to begin building up the nation’s stockpile of cannabis genetics, setting aside half a million dollars to support the work.