IPCARF: improving lncRNA-disease association prediction using incremental principal component analysis feature...

Area of Science:

• Computational biology
• Genomics
• Biomedical informatics

Background:

• Identifying long non-coding RNA (lncRNA)-disease associations is crucial for understanding disease mechanisms and discovering biomarkers.
• Traditional experimental methods struggle to keep pace with the growing volume of biological data for lncRNA-disease association detection.
• Developing effective computational methods is essential for predicting human lncRNA-disease associations.

Purpose of the Study:

• To propose a novel computational algorithm, IPCARF, for predicting lncRNA-disease associations.
• To integrate multiple similarity matrices and machine learning techniques for enhanced prediction accuracy.

Main Methods:

• Developed the IPCARF algorithm combining incremental principal component analysis (IPCA) and random forest (RF).
• Computed disease semantic similarity using a directed acyclic graph and integrated lncRNA similarity and Gaussian nuclear similarity.
• Reduced feature dimensionality using IPCA and trained an RF model for prediction.

Main Results:

• The IPCARF algorithm demonstrated improved prediction performance, achieving an Area Under the Curve (AUC) of 0.8529 before parameter optimization.
• After grid search optimization, the IPCARF algorithm's AUC reached 0.8611.
• IPCARF effectively improved the AUC metric for predicting potential lncRNA-disease associations.

Conclusions:

• The IPCARF method shows superior performance compared to existing prediction methods like LRLSLDA, LRLSLDA-LNCSIM, TPGLDA, NPCMF, and ncPred.
• IPCARF provides a more accurate and efficient approach for predicting lncRNA-disease associations.
• The study highlights the potential of integrated machine learning for advancing lncRNA-disease association research.