A deep learning method for the long-term prediction of plant electrical signals under salt stress to identify salt tolerance

Abstract

Screening salt-tolerant crops at different salt concentrations is of great significance but time-consuming. It remains a major challenge to automatically find the appropriate stress concentration to identify the salt tolerance of crops by using the recorded electrical signals of plants over long periods of time. To solve this problem, we designed a data-driven signal dynamics prediction model (SLSTM-TCNN), based on a one-dimensional convolutional neural network (1D-CNN) combined with a long short-term memory neural network. Furthermore, we developed a quantitative model, named the NaCl stress concentration discrimination model (SCDM), to investigate the relationship between the electrical signals, NaCl stress concentration, and time dependence, and used a salt tolerance classification model (STCM) to discover the most appropriate NaCl stress concentration for distinguishing the salt tolerance of wheat. These methods perform the time-consuming task of selecting salt-tolerant varieties of plants under different NaCl concentrations. The results show that the SLSTM-TCNN could quickly predict the signal dynamics of wheat leaves DeKang961 (salt-tolerant) and Langdon (salt-sensitive) under the ongoing stress of different NaCl concentrations. The accuracy of SCDM for differentiating NaCl stress concentrations increased to 88% and 83%, and that of STCM for classifying salt-tolerant and salt-sensitive varieties reached 92.36%. Finally, it was found that the salt tolerance of the two varieties (DeKang961 and Langdon) was higher when the NaCl concentration was in the range of 50–200 mM. In the future, the method will be a potentially useful tool for identifying the salt tolerance of other crops at the seedling stage.