Deep learning-based cancer disease classification through MicroRNA expression

MicroRNAs (miRNAs) are an RNAs with 22 to 25 nucleotides that are involved in gene regulation by inhibiting the translation or degrading messenger RNA (mRNA). In recent years, finding the associations between miRNA and diseases represents an important area of research. Although several machine learning algorithms have been used to optimize disease classification accuracy value, they remain insufficient with the large and complicated volume of biological data. In this work, we proposed a new disease classification model based on deep learning algorithms. First of all, we convert the miRNA expression of human samples that are affected by cancer disease to image with DeepInsight tools. Then, we will compare three variants of ResNet models for disease classification which are ResNet50, ResNe101, and ResNet152. We consider five cancer disease which are Breast invasive carcinoma (BRCA), Glioma (GBML), Stomach and Esophageal carcinoma (STES), Pan-kidney cohort (KIPAN), and Ovarian Serous Cystadenocarcinoma (OV). Experimental results shows that a combination of DeepInsight and ResNet50 gives the best performance with 92.9% of prediction accuracy, 90.68% as sensitivity, and 94.07% as recall value.