Non-Destructive Sorting of Cannabis Germplasm Based on Predicted Phenotypical Properties Using Spectral Fingerprinting and AI-algorithms
Executive Summary
The ability to non-destructively sort cannabis germplasm based on the predicted phenotypic properties of the resulting plants is an absolute breakthrough in modern breeding and cultivation techniques. It is arguably the biggest breakthrough since the advent of feminized seeds over 40 years ago.
This white paper explores the use of advanced smart camera arrays, including x-ray, RGB (visual spectrum), chlorophyll, multispectral, and hyperspectral imaging systems, coupled with artificial intelligence (AI) algorithms, to provide a comprehensive phenotypic fingerprint of cannabis seeds before they germinate.
By integrating these technologies, breeders and cultivators can predict the traits of future plants with greater accuracy, leading to optimized selection processes and improved consistency in cannabis production.
At Innexo, we are actively pioneering the use and abilities of this amazing technology in cannabis.
Introduction
Cannabis cultivation and breeding have long been reliant on trial-and-error methods to evaluate the potential of individual seeds and plants.
The first revolution (before the 1990’s) of cannabis production came in the form of ‘sinsemillia’ (seedless) production. The focus was to cull all male plants from ‘regular’ seeds in order to have an all-female crop. Although this practice is labor and resource intensive, it is a trusted way of ensuring a non-seeded crop with a stable yield and cannabinoid profile.
The second revolution came in the form of ‘feminized’ seeds. These seeds will express around 99% female plants (XX chromosomes). However, some uncontrolled breeding has led to a certain percentage of varieties being of intersex traits. Even though on a genetic level there is no Y-chromosome. These intersex traits can produce male flowers that contain pollen and can thus lead to unwanted fertilization of the crop, resulting in lower cannabinoid content and yield.
The third revolution came around 10 years ago and consists of a PCR test performed at the seedling stage. This genetic test can positively identify any male plants, before they reveal their sex in a natural way. This saves cultivators and breeders enormous amounts of time, resources and labor. However, to data, this method has been unsuccessful in weeding out the intersex plants, limiting its use to only male/female separation of seedlings.
Right now, we are on the cusp of the forth evolution in early sex and trait determination: spectral fingerprinting.
The advent of non-destructive sorting techniques that utilize spectral fingerprinting allows breeders to predict plant traits before seeds are even germinated.
This technique leverages advanced imaging technologies and machine learning to analyze the physical and chemical properties of seeds, providing valuable insights into their future growth potential.
At Innexo, we are at the forefront of this development. Through the partnership with our sister company, Innoveins Seed Solutions (ISS), we are establishing the future of cannabis breeding, seed production and phenotyping. This conceptual whitepaper lays the groundwork of this research initiative and is meant as an educational tool for all actors in the cannabis industry.
Spectral Fingerprinting: Technologies and Methods
X-Ray Imaging
X-ray imaging is used to analyze the internal structure of cannabis seeds, providing insights into seed viability, health, and potential vigor. This non-destructive technique allows breeders to assess whether seeds contain well-developed embryos and to detect any internal damage that may not be visible externally. X-ray imaging is particularly useful for detecting physical deformities or irregularities that can impact germination success.
RGB (Visual Spectrum) and Chlorophyll Imaging
RGB and chlorophyll imaging systems offer a powerful combination of tools for analyzing the photosynthetic capacity and overall health potential of cannabis germplasm. RGB measures and recognize size, shape and patterns on the outside of the seed. Chlorophyll imaging can predict how efficiently a plant will carry out photosynthesis. Together, these tools help breeders predict growth patterns and biomass production.
Multispectral and Hyperspectral Imaging
Multispectral and hyperspectral imaging systems are essential for capturing detailed spectral signatures of cannabis seeds. These systems detect a wide range of wavelengths (500-2.000 nm) that correspond to specific chemical and physical characteristics, including hormones, proteins, starch, glucose, genetic markers and chromosomes.
By using hyperspectral imaging, breeders could predict the chemical composition of the plants that will emerge from the seeds, helping to select seeds that will produce desired traits, such as sex, color, flowering time, or other agronomical traits.
AI Algorithm Processing for Phenotypic Prediction
The integration of AI-algorithms is a critical component of this non-destructive sorting. Machine learning models are actively being trained to recognize patterns in the spectral data collected from the various imaging systems, allowing for accurate prediction of the phenotypic properties of cannabis plants.
Data Collection, Analysis and Modeling
At Innexo, we perform the controlled grow-out of the analyzed germplasm and annotate the as much traits of each individual plant as possible. This way, the phenotypical data is collected and labeled to each individual seed.
At the end of each grow-out, the phenotypical data collected is combined with the spectral data and fed into the statistical modelling software. This software finds statistical correlations between the two datasets and trains the predictive AI-algorithms.
Predictive models are used to forecast specific phenotypic traits, such as cannabinoid ratios, plant height, yield, and resistance to environmental stressors, etc.
This means that if a trait can be positively identified through statistical spectral correlations, it is possible to train sorting algorithms for this specific trait. With ever more data being collected, the sorting algorithm becomes more and more robust and accurate.
Applications in Cannabis Breeding and Cultivation
Optimized Seed Selection
Non-destructive sorting based on predicted phenotypic traits allows breeders to optimize seed selection by identifying the most viable and promising seeds before they germinate. This results in faster breeding cycles and a higher success rate in producing plants with desired characteristics without going into the field of molecular genetics.
With this technology, breeders can prioritize seeds that are predicted to produce high-yield, cannabinoid-rich plants, significantly reducing the guesswork involved in traditional selection methods.
Consistency in Pharmaceutical Cannabis Production
For pharmaceutical applications, consistency in cannabis production is critical. By employing spectral fingerprinting and AI-based prediction, cultivators can ensure that only seeds likely to produce consistent cannabinoid profiles are selected for production.
This level of control is essential for meeting the stringent quality standards required in the pharmaceutical industry, where variability in cannabinoid content can affect the efficacy and safety of medical products.
This is especially true for the promising creation of F1-hybrids. By combining spectral fingerprinting with these F1-hybrids, off-types can be detected and eliminated from the batch. This leads to even more uniform seeds and plants.
Challenges and Future Directions
Technical Challenges
While the integration of spectral imaging and AI offers significant advantages, there are technical challenges to address. These include the calibration and synchronization of different imaging systems, ensuring that the data collected from each modality is accurate and compatible.
Additionally, the machine learning models require extensive training and validation to ensure their predictions are reliable across different cannabis varieties. At Innexo, we work with trained models from other crops (such as tomatoes, grasses and herbs) and infuse our extensive knowledge on cannabis cultivars and agronomical traits. This will significantly speed up the development process of this groundbreaking technology for the cannabis sector.
Future Research and Development
Trials at Innexo, together with leading seed companies and esteemed breeders, are being conducted to validate the first generation of traits for sorting. In this first wave of ground-breaking development focusses on identifying male and female sex, intersex traits, inflorescence color, flowering time and seed germination upgrading.
Conclusion
The use of non-destructive sorting based on spectral fingerprinting and AI-driven phenotypic prediction represents a significant advancement in cannabis breeding and cultivation.
By leveraging a combination of x-ray, RGB, chlorophyll, multispectral, and hyperspectral imaging technologies, alongside sophisticated and trained AI-algorithms, breeders, seed companies and cultivators can predict plant traits with unprecedented accuracy.
This system holds immense potential for improving seed selection, ensuring consistency in pharmaceutical cannabis production, and streamlining breeding processes.