MedElevator
Войти в исследования ChemMedChem · 2026
Bilingua-DPP-IV: A Bimodal Fusion Convolutional Neural Network Model for Prediction of Dipeptidyl Peptidase IV Inhibitory Peptides.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Zhang Zhixing, Yu Qiule, Yang Chen, Duan Hao, Fang Jiaojiao, Xu Jiahao, Gong Changda, Li Weihua, Liu Guixia, Tang Yun
Аннотация
MOTIVATION: Dipeptidyl peptidase IV (DPP-IV) is a key therapeutic target for type 2 diabetes, and its inhibitory peptides are a promising class of therapeutic agents. However, the discovery of these promising drugs has been restricted, as many existing computational predictors are based on datasets that have remained static for several years.
RESULTS: Here, we first presented a newly curated and expanded benchmark dataset, then introduced Bilingua-DPP-IV, a bimodal deep learning framework that integrates features from both amino acid sequences and molecular structures. Our approach leveraged embeddings from the large-scale pretrained models ESM C and Uni-Mol2, processed through a convolutional neural network with an optimized gated fusion mechanism. On an independent test set, the Bilingua-DPP-IV ensemble model achieved a Matthews correlation coefficient of 0.764 and an area under the receiver operating characteristic curve of 0.963. On an external validation set, the model also demonstrated an accuracy of 0.898, exceeding the performance of existing state-of-the-art predictors, such as StructuralDPPIV.
RESULTS: Here, we first presented a newly curated and expanded benchmark dataset, then introduced Bilingua-DPP-IV, a bimodal deep learning framework that integrates features from both amino acid sequences and molecular structures. Our approach leveraged embeddings from the large-scale pretrained models ESM C and Uni-Mol2, processed through a convolutional neural network with an optimized gated fusion mechanism. On an independent test set, the Bilingua-DPP-IV ensemble model achieved a Matthews correlation coefficient of 0.764 and an area under the receiver operating characteristic curve of 0.963. On an external validation set, the model also demonstrated an accuracy of 0.898, exceeding the performance of existing state-of-the-art predictors, such as StructuralDPPIV.