WheatRPred: deep learning framework for predicting disease resistance protein in wheat and its wild relatives

Wheat, a globally vital staple crop, is under threats from a variety of pathogens, highlighting the urgent need for efficient identification of disease resistance (R) genes. Traditional experimental methods are slow and resource-intensive. To address this, we present WheatRPred, a web-based deep learning framework for accurate classification of R proteins in wheat and its wild relatives. Our integrative approach combines CNN, BiLSTM, and Transformer models which capture spatial, sequential, and contextual features from protein sequences. Trained on a curated dataset of validated R and non-R proteins, the models achieved 95-97% accuracy, outperforming traditional machine learning classifiers. WheatRPred allows users to submit protein sequences and obtain predictions classifying each sequence as resistance or non-resistance, along with confidence scores, thereby accelerating disease resistance research and genomic breeding in wheat. This platform provides a practical resource for researchers and breeders, supporting rapid R gene discovery, marker-assisted selection, and the development of disease-resistant wheat varieties.