Deep learning-based method to identify disease-resistance proteins inOryza sativaand relative species

Abstract

Rice (Oryza sativa) is a significant agricultural crop consumed by more than half of the global population. Its demand is expected to increase due to rising consumption and a growing global population. Moreover, the rice plant is frequently exposed to disease-causing pathogens, such as bacteria, fungi, viruses, and nematodes. Thus, cultivating disease-resistant varieties is an efficient way of disease control compared to pesticide applications. However, the rice plant has a well-defined defense system to prevent the onset of disease, including Pathogen-associated molecular pattern (PAMP)-triggered immunity (PTI) and effector-triggered immunity (ETI). The defense system is controlled by various disease-resistance proteins, such as resistance (R) proteins and pathogen recognition receptors (PRRs). Therefore, the identification of disease-resistance proteins not only reduces the amount of pesticides used in rice fields but also increases their yield. Though some resistant proteins have been characterized, their rapid identification, precise diagnosis, and appropriate management are still lacking. However, few methods based on sequence-similarity and de novo prediction, such as Machine Learning (ML), usually have low prediction power. In this study, we built a state-of-the-art classifier based on Deep Learning (DL) for the early detection of disease-resistance proteins in rice and related species. We compared the DL-based Multi-layer Perceptron (MLP) model with the five well-established ML-based methods using a protein dataset of rice and its related species. The DL-based MLP model outperformed all of the five classifiers on 10-fold cross-validation. The accuracy, Area Under Receiving Operating Characteristic (ROC) curve (AUC), F1-score, precision, and recall were superior in the DL-based MLP model. In conclusion, the MLP model is an effective DL model for predicting disease-resistance proteins with high scores in performance metrics. This study will provide insight to the breeders in developing disease-resistant rice varieties and assist in transforming traditional rice farming practices into a new age of smart rice farming.