Abstract

This study helps medical professionals to find out the genetic information related to five types of cancer. The information stored in RNA is characterized by its abundance, which is responsible for forming proteins in the human body. If there is any harm in this acid, this leads to the formation of cancerous diseases. The RNA contains huge numbers of genes so It is difficult for researchers to isolate the affected genes from the healthy genes. This paper presents a method for clustering the mutated genes for five types of cancer by using a novel feature extraction method with one of the traditional unsupervised machine learning algorithms (Self-Organized Map SOM). The proposed feature extraction method is called the Bi- directional Local-binary-pattern Signal (BLS), it is inspired from three traditional feature extraction methods named the Local Binary Pattern (LBP), Local Line Binary Pattern (LLBP), and Enhanced One-dimensional Local Binary Patterns (EOLBP) to extract the important textures for five types of cancer genes. This method finds the average for both direction features of a specific window in the RNA gene sequence and slides this window along the RNA to extract the features that are the most indicative features of cancerous genes. This study successfully clustered five types of cancer genes using the SOM classifier based on the features extracted by the proposed method. The study used a benchmarked dataset to test the proposed method. The accuracy for distinguishing the muted genes using the proposed method is 99.5%.