TY  - CONF
AU  - Chen, Yaojia
AU  - Niu, Mengting
AU  - Liu, Yi
AU  - Wang, Jiacheng
AU  - Ding, Yijie
AU  - Zou, Quan
ED  - Huang, De-Shuang
ED  - Pan, Yijie
ED  - Zhang, Qinhu
PY  - 2024
DA  - 2024//
TI  - AAHLDMA: Predicting Drug-Microbe Associations Based on Bridge Graph Learning
BT  - Advanced Intelligent Computing in Bioinformatics
SP  - 3
EP  - 14
PB  - Springer Nature Singapore
CY  - Singapore
AB  - Human microbes, intricately intertwined with human hosts, play pivotal roles in drug development and precision medicine by modulating drug efficacy and toxicity. Utilizing microbes in antibacterial development is a new focus, yet understanding their complex interactions with drugs remains a challenge. Identifying microbe-drug associations enhances understanding and accelerates drug development, benefiting research and screening efforts. Given the limitations of biological experiments, there’s a need for computational methods to identify microbe-drug associations, driven by expanding genomic and pharmacological datasets. In this work, we leveraged rich biological information to obtain representation of drugs and microbes, including molecular structure similarity, drug Gaussian interaction profile similarity, drug network topological feature, microbe functional similarity, and microbe sequence feature. Then, the attention-based graph autoencoder was used to learn the latent representations of drugs and microbes. To gather network-level insights, we integrate a novel bridge node concept, which connects drugs and microbes, facilitating the construction of a learnable drug-microbe interaction network. Furthermore, we integrate both attention mechanism and graph convolutional network to identify drug-microbe associations. Experimental results demonstrate that our AAHLDMA model attains an average AUC of 0.9581 and AUPR of 0.9587 through 5-fold cross-validation, consistently outperforming five other state-of-the-art methods. Moreover, case studies corroborate the efficacy of AAHLDMA in identifying potential drug-microbe associations.
SN  - 978-981-97-5692-6
ID  - 10.1007/978-981-97-5692-6_1
ER  -