Transcriptomic Interpretation on Explainable AI-Guided Intuition Uncovers Premonitory Reactions of Disordering Fate in Persimmon Fruit.

Deep neural network (DNN) techniques, as an advanced machine learning framework, have allowed various image diagnoses in plants, which often achieve better prediction performance than human experts in each specific field. Notwithstanding, in plant biology, the application of DNNs is still mostly limited to rapid and effective phenotyping. The recent development of explainable CNN frameworks has allowed visualization of the features in the prediction by a convolutional neural network (CNN), which potentially contributes to the understanding of physiological mechanisms in objective phenotypes. In this study, we propose an integration of explainable CNN and transcriptomic approach to make a physiological interpretation of a fruit internal disorder in persimmon, rapid over-softening. We constructed CNN models to accurately predict the fate to be rapid softening in persimmon cv. Soshu, only with photo images. The explainable CNNs, such as Gradient-weighted Class Activation Mapping (Grad-Class Activation Mapping (CAM)) and guided Grad-CAM, visualized specific featured regions relevant to the prediction of rapid softening, which would correspond to the premonitory symptoms in a fruit. Transcriptomic analyses to compare the featured regions of the predicted rapid-softening and control fruits suggested that rapid softening is triggered by precocious ethylene signal-dependent cell wall modification, despite exhibiting no direct phenotypic changes. Further transcriptomic comparison between the featured and non-featured regions in the predicted rapid-softening fruit suggested that premonitory symptoms reflected hypoxia and the related stress signals finally to induce ethylene signals. These results would provide a good example for the collaboration of image analysis and omics approaches in plant physiology, which uncovered a novel aspect of fruit premonitory reactions in the rapid-softening fate.

Explainable Deep Learning Reproduces a 'Professional Eye' on the Diagnosis of Internal Disorders in Persimmon Fruit.

Recent rapid progress in deep neural network techniques has allowed recognition and classification of various objects, often exceeding the performance of the human eye. In plant biology and crop sciences, some deep neural network frameworks have been applied mainly for effective and rapid phenotyping. In this study, beyond simple optimizations of phenotyping, we propose an application of deep neural networks to make an image-based internal disorder diagnosis that is hard even for experts, and to visualize the reasons behind each diagnosis to provide biological interpretations. Here, we exemplified classification of calyx-end cracking in persimmon fruit by using five convolutional neural network models with various layer structures and examined potential analytical options involved in the diagnostic qualities. With 3,173 visible RGB images from the fruit apex side, the neural networks successfully made the binary classification of each degree of disorder, with up to 90% accuracy. Furthermore, feature visualizations, such as Grad-CAM and LRP, visualize the regions of the image that contribute to the diagnosis. They suggest that specific patterns of color unevenness, such as in the fruit peripheral area, can be indexes of calyx-end cracking. These results not only provided novel insights into indexes of fruit internal disorders but also proposed the potential applicability of deep neural networks in plant biology.