Predicting LncRNA-Disease Association Based on Generative Adversarial Network.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) are increasingly recognized for their roles in human diseases.
• Currently, only a limited number of lncRNA-disease associations are experimentally validated.
• Computational models are crucial for predicting these associations, reducing experimental costs.

Purpose of the Study:

• To develop a predictive model for lncRNA-disease associations using limited known data.
• To propose a novel model, LDA-GAN, based on Generative Adversarial Networks (GANs).

Main Methods:

• LDA-GAN employs Gumbel-softmax technology for differentiable discrete sampling, addressing GAN limitations.
• The model integrates generator and discriminator with a pairwise loss function for optimization.
• Utilizes a semi-supervised learning approach within the GAN framework.

Main Results:

• LDA-GAN demonstrates high stability and efficiency in adversarial learning.
• The model significantly improves prediction accuracy for lncRNA-disease associations.
• Case studies confirm LDA-GAN's ability to identify potential disease associations for lncRNAs.

Conclusions:

• A generative adversarial model, LDA-GAN, is introduced for identifying lncRNA-disease associations.
• LDA-GAN offers an effective computational approach to advance understanding of lncRNA roles in disease.