Integrating machine learning interpretation methods for investigating nanoparticle uptake during seed priming and its biological effects

Abstract

Seed priming by nanoparticles is an environmentally-friendly solution for alleviating malnutrition, promoting crop growth, and mitigating environmental stress. However, there is a knowledge gap regarding the nanoparticle uptake and the underlying physiological mechanism. Machine learning has great potential for understanding the biological effects of nanoparticles. However, its interpretability is a challenge for building trust and providing insights into the learned relationships. Herein, we systematically investigated how the factors influence nanoparticle uptake during seed priming by ZnO nanoparticles and its effects on seed germination. The properties of the nanoparticles, priming solution, and seeds were considered. Post hoc interpretation and model-based interpretation of machine learning were integrated into two ways to understand the mechanism of nanoparticle uptake during seed priming and its biological effects on seed germination. The results indicated that nanoparticle concentration and ionic strength influenced the shoot fresh weight mainly by controlling the nanoparticle uptake. The nanoparticle uptake had a significant slowdown when the nanoparticle concentration exceeded 50 mg L⁻¹. Although other factors, such as zeta potential and hydrodynamic diameter, had no obvious effects on nanoparticle uptake, their biological effects cannot be ignored. This approach can promote the safer-by-design strategy of nanomaterials for sustainable agriculture.