POCALI: Prediction and Insight on CAncer LncRNAs by Integrating Multi-Omics Data with Machine Learning

Area of Science:

• Genomics and Bioinformatics
• Cancer Research
• Molecular Biology

Background:

• Long non-coding RNAs (lncRNAs) are emerging as crucial biomarkers for cancer detection and treatment.
• Current computational methods for identifying cancer lncRNAs often lack comprehensive multi-omics integration and systematic feature contribution analysis.

Purpose of the Study:

• To develop and validate POCALI, an algorithm for identifying cancer lncRNAs by integrating extensive multi-omics features.
• To systematically evaluate the predictive contribution of different omics features in cancer lncRNA identification.

Main Methods:

• Developed POCALI, an algorithm integrating 44 omics features across six categories.
• Explored feature contributions to cancer lncRNA prediction, including individual feature impact.
• Benchmarked POCALI against existing methods and validated novel predictions against cancer phenotypes and genomics.

Main Results:

• POCALI identified secondary structure and gene expression features as strong predictors, with epigenomic features as moderate predictors.
• POCALI demonstrated superior performance, particularly in sensitivity, and identified a greater number of candidate cancer lncRNAs compared to other methods.
• Novel cancer lncRNAs predicted by POCALI exhibited significant associations with cancer phenotypes, mirroring known cancer lncRNAs.

Conclusions:

• POCALI effectively identifies novel cancer lncRNAs by integrating multi-omics data.
• The study provides insights into the multifaceted contributions of various omics features to cancer lncRNA prediction.
• This work facilitates the discovery of new cancer biomarkers and enhances our understanding of lncRNA roles in cancer.