BiGAN: LncRNA-disease association prediction based on bidirectional generative adversarial network

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) play critical roles in human physiological processes and hormone regulation.
• Mutations in lncRNAs are implicated in various human diseases.
• Predicting lncRNA-disease associations is vital for understanding disease mechanisms and aiding diagnosis.

Purpose of the Study:

• To develop a computational model for predicting novel long non-coding RNA (lncRNA)-disease associations.
• To leverage multiple similarity features for enhanced prediction accuracy.

Main Methods:

• A bidirectional generative adversarial network (BiGAN) model was developed, comprising an encoder, generator, and discriminator.
• Features were constructed using disease semantic similarity, lncRNA sequence similarity, and Gaussian interaction profile kernel similarities.
• Latent features were mapped to predict unverified lncRNA-disease associations.

Main Results:

• The BiGAN model significantly outperformed existing state-of-the-art methods in cross-validation.
• The model successfully predicted candidate lncRNAs for renal and colon cancers.
• Case studies confirmed that approximately 70% of top-ranked predicted lncRNAs are supported by recent biological research.

Conclusions:

• The proposed BiGAN model exhibits accurate predictive capabilities for lncRNA-disease associations.
• This computational approach offers a valuable tool for exploring lncRNA functions in disease contexts.