Fig. 6

Schematic diagram of the established in silico screening system of ROS modulators. The screening comprised three interconnected phases that allowed the identification of a variety of ROS modulators. The first phase developed a ROS inference system based on DNN learning that combines ROS production data in plant cells and multidimensional chemical features of compounds (a). The second phase was in silico first screening of seven million commercially available compounds using the ROS inference model (b). The third phase was in silico second screening of selected compounds by visualizing the chemical space using the GTM method (b). This system is based on in silico analysis and has advantages in terms of both selection time and cost compared to actual measurements in plant cells