Tumor classification and biomarker discovery based on the 5'isomiR expression level

Area of Science:

• * Molecular Biology
• * Bioinformatics
• * Oncology

Background:

• * MicroRNA isoforms (isomiRs) play a significant role in gene regulation, similar to canonical microRNAs.
• * IsomiRs have demonstrated potential in discriminating between various cancer types.
• * Identifying key isomiRs is crucial for advancing cancer understanding and treatment.

Purpose of the Study:

• * To develop a method for identifying reliable sets of cancer-associated 5'isomiRs.
• * To utilize these isomiRs for multiclass tumor classification.
• * To leverage genetic algorithms (GA) and Random Forest (RF) for this classification task.

Main Methods:

• * Construction of a hybrid GA-RF model.
• * Application to TCGA isomiR expression data for multiclass tumor classification.
• * Feature selection and importance analysis of identified 5'isomiRs.

Main Results:

• * Identification of 100 optimal feature sets, each containing 50 5'isomiRs.
• * Achieved an average sensitivity of 92% in classifying samples from 32 tumor types.
• * Functional enrichment analysis revealed involvement of top isomiRs in transcription activation, protein kinase activity, and cell-cell adhesion.

Conclusions:

• * 5'isomiRs show promise as biomarkers for multiclass tumor classification.
• * The GA/RF model effectively performs tumor classification using independent sets of optimal predictor 5'isomiRs.
• * This approach enhances our understanding of isomiR roles in cancer.