Rasch analysis of chromosomal events as indicators of colorectal tumor severity

Abstract

A Single Nucleotide Polymorphism (SNP) is used to estimate latent tumor severity of colorectal cancer. The Rasch model is applied to a data set of 78 colorectal tumors and 727 cytobands, in order to obtain latent tumor severity estimates and cytoband information estimates. The origin of this project lies in a predominant finding in cancer (including colon cancer) research of the last four decades. Namely, the simultaneous occurrence of cancer with chromosomal abnormalities. The value of tumor severity analysis using Rasch technology lies in the information it can provide to the medical professional with regard to the severity of the tumor using only a biopsy. Hence, this method could be a valuable tool for preoperative staging. The parameter estimates are obtained using maximum likelihood estimation and the appliance of a penalty in the log-likelihood. The penalty technique added to the Rasch model is valuable in its ability to identify model parameter estimates for tumors without chromosomal events on the cytobands. However, too high penalty values lead to non-convergence of model parameter estimates. The interpretation of tumor severity parameter estimates is similar to tumor grades given by medical experts. Besides information on the severity of specific tumors, cytoband estimates give insight to the extent of the relation between chromosomal aberrations on a specific cytoband location and the severity of a colorectal tumor.