Identification of Human RNA m5C sites using CapsuleNet Architecture

An important post-transcriptional alteration is 5-methylcytosine (m5C). It has become more prevalent as technology has advanced in many different RNAs. Previous research has shown that m5C is crucial for many biological processes, including tRNA recognition, RNA stability and the metabolic process of RNA. Traditional detection processes are expensive and time-consuming. As a result, computer models are frequently employed to locate the m5C locations. It is possible to create an effective prediction model because of deep learning’s extensive computing capabilities. Therefore, developing a reliable computational method for predicting RNA modification sites is essential. In this work, we have proposed a CapsuleNet-based architecture for the classification of m5C sites. The one-hot technique is used to convert biological sequences into numerical data. The CNN model is then given the numerical data in order to extract the concealed information. Further, this concealed information is shared with the CapsuleNet framework to get the m5C sites classified. The proposed architecture is evaluated on the human m5C dataset. CapsuleNet-based architecture has demonstrated improved results when compared with already available predictors.