Recent researches of Colorectal Cancer (CRC) aim to look for the answers for its occurrence in the disrupted gene expressions by examining colorectal carcinogenic and healthy tissues with different microarray technologies. In this paper, we propose a novel generative modelling of the Bayes’ classification for the CRC problem in order to differentiate between colorectal cancer stages. The main contribution of this paper is the solution of the distinguishing problem between the critical CRC stages that remained unsolved in the published materials - distinguishing the stage I with stage IV, and stage II with stage III. The Bayesian classifier enabled application of the ’smoothing procedure’ over the data from the third stage, which succeeded to distinguish the probabilities of the mentioned stages. This results are obtained as a continuation of our previous work, where we proposed methodologies for statistical analysis of colorectal gene expression data obtained from the two widely used platforms, Affymetrix and Illumina. Furthermore, the unveiled biomarkers from the two platforms were used in our generative approach for modelling the gene expression probability distribution and were used in the Bayes’ classification system, where we performed binary classifications. This novel approach will help in producing an accurate diagnostics system and precising the actual stage of the cancer. It is of great advantage for early prognosis of the disease and appropriate treatment.
Microarray Analysis; Machine Learning; Bayes’ Theorem; Colorectal Cancer; Classification