Exploring Insecticidal Molecules with Random Forest: Toward High Insecticidal Activity and Low Bee Toxicity

Insecticidal molecules with high activity are crucial for global pesticide reduction and food security. However, their usage is limited by their concomitant high toxicity to bees. Balancing insecticidal activity and bee toxicity remains a critical challenge in the exploitation of new insecticidal molecules. In this study, we propose a novel strategy for exploiting molecules that are both highly effective against pests and minimally harmful to bees. A series of molecules were synthesized and tested to train a machine learning (ML) model for predicting insecticidal activity against pests. Meanwhile, another ML model was trained by using publicly available data to predict bee toxicity. The models demonstrated good performance, with mean AUC values of 0.88 ± 0.05 for insecticidal activity and 0.91 ± 0.01 for bee toxicity. By integrating these two models, we successfully predicted and experimentally validated a molecule that exhibited a high insecticidal activity and low bee toxicity. This dual-ML-model approach offers a promising pathway for the development of insecticidal molecules that are both effective and environmentally safe, thereby contributing to sustainable agricultures.

Pesticide Exposure and Effects on Non-Apis Bees

Bees are essential pollinators of many crops and wild plants, and pesticide exposure is one of the key environmental stressors affecting their health in anthropogenically modified landscapes. Until recently, almost all information on routes and impacts of pesticide exposure came from honey bees, at least partially because they were the only model species required for environmental risk assessments (ERAs) for insect pollinators. Recently, there has been a surge in research activity focusing on pesticide exposure and effects for non-Apis bees, including other social bees (bumble bees and stingless bees) and solitary bees. These taxa vary substantially from honey bees and one another in several important ecological traits, including spatial and temporal activity patterns, foraging and nesting requirements, and degree of sociality. In this article, we review the current evidence base about pesticide exposure pathways and the consequences of exposure for non-Apis bees. We find that the insights into non-Apis bee pesticide exposure and resulting impacts across biological organizations, landscapes, mixtures, and multiple stressors are still in their infancy. The good news is that there are many promising approaches that could be used to advance our understanding, with priority given to informing exposure pathways, extrapolating effects, and determining how well our current insights (limited to very few species and mostly neonicotinoid insecticides under unrealistic conditions) can be generalized to the diversity of species and lifestyles in the global bee community. We conclude that future research to expand our knowledge would also be beneficial for ERAs and wider policy decisions concerning pollinator conservation and pesticide regulation.

A spatially explicit model of landscape pesticide exposure to bees: Development, exploration, and evaluation

Pesticides represent one of the greatest threats to bees and other beneficial insects in agricultural landscapes. Potential exposure is generated through compound- and crop-specific patterns of pesticide use over space and time and unique degradation behavior among compounds. Realized exposure develops through bees foraging from their nests across the spatiotemporal mosaic of floral resources and associated pesticides throughout the landscape. Despite the recognized importance of a landscape-wide approach to assessing exposure, we lack a sufficiently-evaluated predictive framework to inform mitigation decisions and environmental risk assessment for bees. We address this gap by developing a bee pesticide exposure model that incorporates spatiotemporal pesticide use patterns, estimated rates of pesticide degradation, floral resource dynamics across habitats, and bee foraging movements. We parameterized the model with pesticide use data from a public database containing crop-field- and date-specific records of uses throughout our study region over an entire year. We evaluate the model performance in predicting bee pesticide exposure using a dataset of pesticide residues in pollens gathered by bumble bees (Bombus vosnesenskii) returning to colonies across 14 spatially independent landscapes in Northern California. We applied alternative model formulations of pesticide accumulation and degradation, floral resource seasonality, and bee foraging behavior to evaluate different levels of detail for predicting observed pesticide exposure. Our best model explained 73 % of observed variation in pesticide exposure of bumble bee colonies, with generally positive correlations for the dominant compounds. Timing and location of pesticide use were integral, but more detailed parameterizations of pesticide degradation, floral resources, and bee foraging improved the predictions little if at all. Our results suggest that this approach to predict bees' pesticide exposure has value in extending from the local field scale to the landscape in environmental risk assessment and for exploring mitigation options to support bees in agricultural landscapes.

Insecticide application timing effects on alfalfa insect communities

Timing of insecticide application can impact efficacy, given variation in both weather and development of the crop and its insect pests. Both target and nontarget insects may vary in life stage and abundance at the time of application. In alfalfa Medicago sativa L. cropping systems, producers have interest in early-season insecticide applications to eliminate last-minute decisions about preharvest applications for alfalfa weevil Hypera postica (Gyllenhal) (Coleoptera: Curculionidae). The standard recommendation is based on scouting larvae close to the first harvest time. We compared early and standard timing of application of a lambda-cyhalothrin pyrethroid on alfalfa pest and beneficial insects. Field trials at a university research farm were conducted in 2020 and 2021. In 2020, early application was as effective as the standard timing against alfalfa weevil, as compared to the untreated control, but less effective than the standard timing in 2021. Effects of timing against Lygus bugs (Hemiptera: Miridae), grasshoppers (Orthoptera: Acrididae), and aphids (Hemiptera: Aphididae) were inconsistent between years. We observed the potential for early application to reduce negative impacts on ladybird beetles (Coleoptera: Coccinellidae) and spiders (Araneae), however, damsel bugs (Hemiptera: Nabidae) were similarly reduced by insecticide application regardless of timing. Overall arthropod community composition differed by both year and treatment. Future research should explore potential trade-offs of spray timing at larger spatial scales.