Application of machine learning for herbicide characterization

Herbicides are chemical molecules used for destruction of weeds. Massive usage of herbicides has resulted in two global problems: increase in weed resistance and harmful impact of human health [1, 2]. In order to facilitate development of novel, more specific herbicides and of strategies for impeding the weed resistance, we have carried out extensive in silico analysis of the set of herbicides. Herein, we present results revealing links between structural, physicochemical, ADME (Absorption, Distribution, Metabolism, Excretion) and toxic features for herbicides (Figure 1). The analysis has been done by using proper machine learning approaches. References: [1] A. Forouzesh, E. Zand, S. Soufizadeh, S. S. Foroushani, Weed Res. 55 (2015) 334-358. [2] V. I. Lushchak, T. M. Matviishyn, V. V. Husak, J. M. Storey, K. B. Storey, EXCLI J. 17 (2018) 1101-1136.