An approach to forecast human cancer by profiling microRNA expressions from NGS data

Area of Science:

• Biochemistry and Molecular Biology
• Bioinformatics and Computational Biology
• Oncology

Background:

• MicroRNAs (miRNAs) are short, non-coding RNA molecules crucial for post-transcriptional gene regulation.
• Aberrant miRNA expression is linked to various diseases, including cancer, making them potential biomarkers.
• miRNA expression profiling helps identify disease-specific expression level differences, aiding disease progression understanding.

Purpose of the Study:

• To develop an effective cancer prediction system using machine learning techniques applied to miRNA expression data.
• To explore miRNA expression profiling and normalization methods for disease prediction.
• To build a classifier using Support Vector Machines (SVM) for cancer type identification.

Main Methods:

• Utilized Next-Generation Sequencing (NGS) data for miRNA expression profiling.
• Applied machine learning, specifically Support Vector Machines (SVM), for classification.
• Trained and validated the model using data from hepatocellular carcinoma, bladder cancer, and lung cancer, employing 10-fold cross-validation.

Main Results:

• Achieved high prediction accuracies: 97.56% for lung cancer, 97.82% for hepatocellular carcinoma, and 95.0% for bladder cancer.
• Validated the system with separate test sets, consistently showing prediction accuracies above 90%.
• Identified significant miRNAs for cancer prediction through differential expression ranking.

Conclusions:

• miRNA expression profiling is an effective method for disease identification, particularly cancer.
• The accuracy of prediction is dependent on the availability of a sufficiently large and well-curated database.
• Machine learning approaches, combined with miRNA profiling, offer a promising avenue for developing robust diagnostic and prognostic tools.