Circular RNA-Drug Association Prediction Based on Multi-Scale Convolutional Neural Networks and Adversarial Autoencoders

The prediction of circular RNA (circRNA)-drug associations plays a crucial role in understanding disease mechanisms and identifying potential therapeutic targets. Traditional methods often struggle to cope with the complexity of heterogeneous networks and the high dimensionality of biological data....

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:International journal of molecular sciences Ročník 26; číslo 4; s. 1509
Hlavní autoři: Wang, Yao, Lei, Xiujuan, Chen, Yuli, Guo, Ling, Wu, Fang-Xiang
Médium: Journal Article
Jazyk:angličtina
Vydáno: Switzerland MDPI AG 11.02.2025
MDPI
Témata:
Algorithms
Autoencoder
Case studies
Computational Biology - methods
Convolutional Neural Networks
Deep learning
Drugs
Health aspects
Humans
Information sources
MicroRNAs
Neural networks
Neural Networks, Computer
Pharmaceutical Preparations - chemistry
Proteins
RNA
RNA, Circular - chemistry
RNA, Circular - genetics
RNA, Circular - metabolism
Taiwan
adversarial autoencoder
circular RNA-drug association prediction
multi-scale convolutional neural network
ISSN:1422-0067, 1661-6596, 1422-0067
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The prediction of circular RNA (circRNA)-drug associations plays a crucial role in understanding disease mechanisms and identifying potential therapeutic targets. Traditional methods often struggle to cope with the complexity of heterogeneous networks and the high dimensionality of biological data. In this study, we propose a circRNA-drug association prediction method based on multi-scale convolutional neural networks (MSCNN) and adversarial autoencoders, named AAECDA. First, we construct a feature network by integrating circRNA sequence similarity, drug structure similarity, and known circRNA-drug associations. Then, unlike conventional convolutional neural networks, we employ MSCNN to extract hierarchical features from this integrated network. Subsequently, adversarial characteristics are introduced to further refine these features through an adversarial autoencoder, obtaining low-dimensional representations. Finally, the learned representations are fed into a deep neural network to predict novel circRNA-drug associations. Experiments show that AAECDA outperforms various baseline methods in predicting circRNA-drug associations. Additionally, case studies demonstrate that our model is applicable in practical related tasks.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:1422-0067
1661-6596
1422-0067
DOI:10.3390/ijms26041509