Modeling and Optimization of Synergistic Interactions

Understanding the bioactivities of mixtures, including botanical extracts, formulations, and tank mixtures, is challenging due to the complex interactions among components. I evaluate conventional reference (null) models, such as Loewe additivity, Bliss independence, HSA, and their extensions, to determine their applicability in predicting the mixture toxicity of insecticides and botanicals. Recognizing their limitations in capturing intermolecular interactions, I also explore parametric synergy models like BRAID, MuSyC, and Zimmer models, which provide greater precision in quantifying dose-dependent interactions but require extensive experimental data and often yield conflicting parameter estimates. To improve predictive accuracy, I apply mixture experimental design, developing mathematical models that capture nonlinear interactions in mixtures of botanicals. Additionally, I use single- and multi-objective optimization algorithms to fine-tune the blending ratios, ensuring the balance among efficacy, ecotoxicity, and phytotoxicity. These data-driven approaches provide a reliable framework for designing highly effective, selective, and sustainable mixtures of insecticides and botanicals for pest management.

Computer vision for studying insect behaviors

Behavioral studies often provide crucial evidences in chemical ecology and toxicology. However, conventional physical marking methods (stickers, paints, dyes) can alter insect behavior, unsuitable for small species, and limited to basic metrics like distance moved, velocity, and time spent in specific areas.

To overcome these limitations, I employ computer vision techniques, including convolutional neural networks and contour detection, to automate behavioral analysis in a non-invasive, detailed, and quantifiable manner. This approach enables measurement of subtle behaviors such as antennal movement, leg angles, and body tilting, which were previously challenging to analyze. Additionally, many toxicological symptoms, such as convulsions, tremors, and prostration, lack standardized quantification. By applying computer vision, I redefine these terms into measurable metrics, transforming them into robust toxicological evidence. Furthermore, I integrate classification and clustering algorithms to analyze insect responses to novel compounds. This approach enables me to predict modes of action (article submitted), synergistic mechanisms between insecticides (article in preparation), insecticide resistance mechanism, repellency, attraction, and host selection.

Testing guidelines for biocides and repellents

Working across these areas has led me to develop an interest in the regulatory aspects of pest management. I have contributed to governmental testing guidelines for biocidal products and participated in developing a consumer information website on pest control products. My works include:

1) Participating in the formulation of official protocols for evaluating insecticidal and repellent products. To improve affordability and reproducibility, I designed testing materials to be 3D-printable, ensuring accessibility across different laboratories.

2) Helping develop a consumer information website, which includes safety guidelines and a whistleblowing system for reporting product issues.

3) Assessing the legal status of synergists and natural products, contributing to discussions on their regulation and approval.