308 – SPEED: Statistics for Health and Health Policy
Predicting Class Membership using Imputation Methods for Clinical Variable for Hepatocellular Carcinoma (HCC)
Amrina Ferdous
Washington State University
Nairanjana Dasgupta
Washington State University
Sayed Daoud
Washington State University
It is an accepted fact that infection with Hepatitis C virus (HCV) is a leading risk factor for chronic liver disease progression, including cirrhosis and hepatocellular carcinoma(HCC). Prevalence rates of HCV infection range from 2% in the US to as high as 14% in developing countries such as Egypt. In an attempt to define the molecular signatures of HCV-induced HCC, the methylation signatures in Egyptian tissue samples from patients with active HCC, patients with HCV and normal liver tissues were compared using a panel of genes that are commonly hyper methylated in other solid tumors. The prognostic impact of the aberrant promoter methylation status of genes was also correlated to the clinicopathological parameters of patients. However, in most cases, clinic-pathological variables are not available for normal patients. This kind of situation is common in diagnostic study of cancer. What we did was solve this problem our two step algorithm for data imputation. Using our imputation methods the probability of correct class prediction increased from 60% to 72%. R-square has also increased. This approach can help other scientists as who are facing the problem of missing data as well.