Prediction of ncRNA-Disease Association Based on Correntropy Induced Loss Matrix Factorization Model

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Human diseases are closely linked to non-coding RNA (ncRNA) regulation.
• Accurate prediction of ncRNA-disease associations is crucial for disease diagnosis, treatment, and prevention.
• Existing algorithms often exhibit suboptimal performance in identifying these associations.

Purpose of the Study:

• To develop an advanced algorithm for predicting ncRNA-disease associations.
• To improve upon the performance of current prediction methods.

Main Methods:

• Developed a Matrix Factorization method incorporating a Correntropy Induced Loss (C-loss) function (C-lossMF).
• Constructed ncRNA and disease similarity matrices, extracting key information.
• Employed matrix decomposition on the ncRNA-disease association matrix using L2 and C-loss.
• Integrated collaborative regularization from similarity matrices.
• Utilized a combined semi-quadratic optimization and gradient descent approach for model optimization.

Main Results:

• The C-lossMF model demonstrated superior performance compared to other advanced models.
• Evaluated using five-fold cross-validation on four distinct datasets.
• Achieved higher accuracy in predicting ncRNA-disease associations.

Conclusions:

• C-lossMF offers a significant improvement in predicting ncRNA-disease associations.
• The method effectively leverages similarity information for enhanced prediction.
• This advancement holds potential for improved disease analysis and therapeutic strategies.