A Whole New World: Using AI to Identify Biological Control Agents
Authored By: Dr. Mikael Courbot, Chief Technology Officer at Micropep and Jean-Claude Haw-King-Chon Director of Data Science & Information Technology
Micropep is leading the charge when it comes to using AI to enable the discovery of biocontrol agents. In this blog, we’ll discuss how and why we use AI for the discovery of these new tools.
We need new options because the power of current pesticides is failing: Weeds, insects, and fungi are developing resistance to conventional approaches, while the number of hungry mouths to feed is increasing. Simultaneously, few crop protection products are being brought to market to replace defunct ones. In fact, since 1984, companies have only introduced one herbicide with a new mode of action, while herbicide-resistant weeds increased more than 600% during that time. Similarly, over 500 species of spiders, mites, and insects have developed some resistance against insecticides. There’s no denying it, farmers’ ability to grow enough food to meet global demand is threatened if we can’t find new solutions. The global population is increasing, making the problem even more urgent. Moreover, companies must bring new products to market while working within an increasingly stringent regulatory landscape.
Here’s where AI shines: It can accelerate scientists’ ability to sift through data in a time- and cost-effective manner, identify patterns, and uncover new compounds that support crop health and meet regulatory expectations, with an eye for those more environmentally friendly because they have fewer off-target effects. At Micropep, we think that artificial intelligence (AI) will be instrumental in meeting a growing need for new agricultural options in the form of environmentally friendly biocontrol agents — biological compounds that support plant health in a way that is complementary to pesticides.
How AI Aids in Discovery
In light of regulatory pressure due to environmental and societal concerns, scientists need to think beyond the classic small-molecule chemical solutions. True, AI can enable scientists to identify new synthetic pesticides. On the other hand, it can also help identify biocontrol agents, an attractive, additional solution for crop protection. Using AI to sift through naturally derived molecules (that modulate plant health and defend against biotic and abiotic stresses) opens the door to identifying a greater number of candidate molecules which are increasing the plant defense against biotic and abiotic stresses and potentially having less environmental impact.
AI revolutionizes agricultural research by harnessing diverse technologies to analyze vast datasets. Traditional ML (machine learning) techniques, such as feature extraction and classification, are instrumental in plant disease detection. Algorithms like c4.5 and linear support vector machines distinguish between healthy and diseased plants, though they require ample annotated data for training. DL (deep learning) technologies like CNNs (convolutional neural network) and DBNs (deep belief network) offer promising avenues for identifying plant abnormalities, automatically learning features that traditional methods may overlook. However, DL models demand significant labeled data and computational resources. Furthermore, CV (computer vision) algorithms, including object detection and semantic segmentation, enhance disease symptom identification, yet they, too, necessitate extensive labeled datasets.
By integrating ML, DL, and CV, AI expedites the discovery process by identifying patterns and virtually screening numerous candidates, surpassing human capacity. This accelerates lead candidate identification in agricultural research, ushering in a new era of efficiency.
Applying AI to Identify Micropeptides
At Micropep, we believe that micropeptides — short 10-20 amino acid sequences that modify cell biology to regulate the phenotype of a target organism in a highly specific manner — are the answer to developing the next generation of efficient and sustainable solutions for crop protection. Our platform combines the power of AI and computational biology to browse genomes and compile libraries of potential micropeptides for future products. The candidate libraries are screened with other AI approaches, including machine learning to predict their properties and deep learning to predict how micropeptides and their target proteins interact. In parallel, Micropep uses generative AI to produce novel micropeptide sequences. In this application, we use deep learning to predict those that will bind strongly to target proteins to determine which predicted micropeptides are most likely to be active.
Subsequently, our AI algorithms help us efficiently predict and prioritize candidates according to our predefined set of criteria. When choosing our criteria, we aim to find the best micropeptide sequences with superior high efficacy, safety, stability, and excellent manufacturing potential, all of which will contribute to identifying a new active ingredient optimally designed to support plant health. Ultimately, our system enables us to narrow the number of candidates from trillions to just a few hundred, which are then screened further through AI-powered in silico screening assays. Those few that show promise in silico are subsequently tested in efficacy assays on actual plants. Using this system, Micropep has demonstrated the acceleration in the discovery of new micropeptides while keeping screening costs down.
The results have been bountiful. Thanks in part to our first generation of AI algorithms, we have several biocontrol agents with different modes of action in our pipeline, including our lead candidate, Promisin, a micropeptide that controls the growth of specific fungi by disrupting their cell membranes. Moreover, in using AI to discover new micropeptides, Micropep and our collaborators are generating a massive body of data to advance the cumulative knowledge of the relatively new field of micropeptide research forward considerably.
Forging a New Path Forward with AI
AI platforms enable us to derive insights from vast data that could otherwise take humans years or even a lifetime to process. By using AI, Micropep scientists are speeding up the process of finding new potential biocontrol agents to discover more environmentally friendly solutions for crop protection. In our efforts, we are seeking out new micropeptides that limit the growth of weeds, fungi, and other pests in a highly targeted manner, as well as micropeptides that promote crop health. Through these efforts, we strongly believe we will bring new earth-friendly products to market in a couple of years that promote maximum crop yields. Ultimately, our goal is to help farmers meet the growing global demand for food in the most sustainable way possible.