SDLDA: lncRNA-disease association prediction based on singular value decomposition and deep learning.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) are crucial regulators of biological processes and human diseases.
• Experimental verification of lncRNA-disease associations is limited due to high costs.
• Existing computational methods struggle to model complex lncRNA-disease relationships.

Purpose of the Study:

• To develop an advanced computational framework for predicting lncRNA-disease associations.
• To overcome limitations of traditional matrix factorization methods.
• To improve the accuracy and efficiency of identifying potential lncRNA-disease links.

Main Methods:

• Proposed a hybrid computational framework named SDLDA.
• Utilized singular value decomposition (SVD) for linear feature extraction.
• Employed deep learning for non-linear feature extraction.
• Combined linear and non-linear features for robust prediction.

Main Results:

• SDLDA demonstrated superior performance compared to existing methods in leave-one-out cross-validation.
• The integration of linear and non-linear features enhanced predictive power.
• Case studies successfully validated 28 out of 30 cancer-related lncRNAs.

Conclusions:

• The hybrid SDLDA framework effectively predicts lncRNA-disease associations.
• SDLDA offers a more powerful approach than methods using only matrix factorization or deep learning.
• This tool can accelerate the discovery of disease-related lncRNAs and advance understanding of disease mechanisms.