Artificial Intelligence is now being used to ensure that plants modified through gene editing are not only successful in the laboratory but can also perform successfully under field conditions, meaning that they are truly robust, productive, and adapted to real-world agricultural growing conditions.
This refers to the journey between the moment scientists modify a plant’s DNA in the laboratory and the moment that modification produces concrete and measurable results under real-world conditions. Scientists use artificial intelligence algorithms and predictive models to determine the best way to cross or select plants.
Indeed, in the laboratory, a genetic modification (using CRISPR, for example) may appear ideal. However, and this is the central challenge, once the plant is grown in a real field and exposed to variations in climate, soil conditions, or diseases, its actual performance is often lower than predicted. This difference between laboratory results and field performance is known as the performance gap.
AI is specifically used to analyze and reduce this gap, helping to ensure that genetically modified plants perform just as effectively in natural environments as they do in the laboratory.
The New Frontier of Gene Editing
The next frontier in crop gene editing is understanding how edited traits interact with the rest of the genome under field conditions, a task for which AI is increasingly being used. AI helps breeders predict how genetically modified traits will perform in the field.
Gene editing opens the door to targeted modifications in crops, but the genetic background in which a modification is expressed can determine whether it succeeds or fails in the field. Genetic background refers to the entire set of hereditary material – the genome – encoded in the DNA of a plant’s cells. It is the biological program that determines how the plant functions, grows, develops, and expresses its characteristics. These elements can influence the performance of a new genetic modification.
However, this concept has not yet been fully exploited because researchers still do not completely understand how different genetic backgrounds influence crop characteristics. This is where AI comes into play.
The Major Contributions of AI
Artificial Intelligence is transforming gene editing technologies such as CRISPR-Cas9, shifting them from an empirical trial-and-error approach toward precision molecular engineering. AI plays four major roles:
- Maximizing cutting precision (guide RNA design).
To modify a gene, scientists must create a “guide RNA” that directs the molecular scissors (Cas9) to the exact target location. AI analyzes billions of sequences to predict which guide will be the most effective while minimizing off-target effects, meaning unintended cuts at incorrect locations in the DNA. - Predicting cellular repair.
Once the DNA has been cut, the cell attempts to repair it on its own. Deep Learning models (such as inDelphi or CRISPR-Spread) can predict in advance the exact sequence that the cell will produce during the repair process, making it possible to anticipate the final result without waiting for laboratory tests. - Designing new enzymes (Generative AI).
Instead of simply reusing naturally occurring enzymes, generative models (similar to ChatGPT but trained on proteins) can create synthetic molecular scissors that are smaller, more precise, or capable of operating in regions of DNA that were previously inaccessible. - Predicting effects on the organism (from laboratory to field).
Modifying a gene can trigger a chain of effects. In agriculture or medicine, AI can simulate the interaction between a genetic modification and its environment (climate, soil, diseases, etc.) to help ensure that the edit produces the desired outcome under real-world conditions.
Many seed companies are exploring the potential of AI in gene editing, particularly for crops such as rapeseed, GABA tomatoes, and non-browning bananas. The combined use of predictive breeding and gene editing could make it possible to develop new varieties at half the cost and in half the time required by traditional breeding methods.
