Function Prediction of Disease-Related Long Intergenic Non-Coding RNA Using Random Forest

Abstract

A non-coding RNA (ncRNA) is an RNA molecule that is not coded into a protein. ncRNAs are important regulators of gene expression, and have a wide range of functions in cellular and developmental processes. However, to date, only a few well-studied ncRNAs have provided important clues about their relevance to particular diseases.This study aims to find disease-related mRNAs and long intergenic non-coding RNAs (lincRNAs) from microarray data. More specifically, this study focuses on acute lung injury (ALI) disease. It seeks to distinguish between ALI and normal one. We utilize microarray data including mRNAs and lincRNAs that contain transcriptomic characterization of mice's brain inflammation during ALI.These samples are obtained at six time points (1h, 1.5h, 3h, 4h, 18h, and 24h) after injection of oleic acid or physiological serum. Because microarray data contains a large number of genes with a small number of samples, we must extract genes that are relevant for classification. First, we divide the dataset into three groups according to elapsed time after injection: group1 with early elapsed time (1h, 1.5h, 3h, and 4h), group2 with late elapsed time (18h and 24h), and group3 with full-length elapsed time (1h to 24h). We then extract some genes using a random forest algorithm for each group. A random forest algorithm can identify which features (genes) are important for classification. Finally, we obtain some important genes (mRNAs and lincRNAs) for classification as output.From these result, we distinguish between mRNA and lincRNA, and extract some mRNAs and lincRNAs using a random forest algorithm from microarray data. The final results indicate that some extracted lincRNAs may function like some mRNAs in ALI disease. This study provides new knowledge for biologists and medical scientists by investigating lincRNAs that may be related to ALI disease.