Identifying Markers Of Hepatocellular Carcinoma For Early Diagnosis Based On The Relative Expression Of Genes

Hepatocellular carcinoma (HCC) is one of the most common and aggressivecancers. It is the second cancer killer in China. Liver cirrhosis is the underlying diseasein>80%of HCC cases. Lacking of effective earlier diagnosis biomarker in cirrhoticpatients leads to high mortality of Hepatocellular carcinoma. Currently, small equivocalliver lesions detected by imaging techniques need biopsy confirmation in surveillancefor hepatocellular carcinoma. Identifying markers of hepatocellular carcinoma for earlydiagnosis based on the gene expression data has its practical significance. At presentmany algorithms have been developed for construct HCC diagnosis classifier, such asrandom forest, logistic regression model. Unfortunately, many classifiers have failed tovalidate by other methods or in new cohorts of patients. To reduce the impact ofdetection batch effects on classification results, we proposed a method to identifyprecancerous lesions and early hepatocellular carcinoma based on the relative geneexpressions of cirrhosis tissue and hepatocellular carcinoma tissue. Based on twodatasets for HCC and cirrhosis, we extract genes that are dysregulated in patients withHCC versus patients with cirrhosis and identify42significant HCC related GO termsincluding “cell cycle”;” cell proliferation”;” cell division”;” DNA replication” etc. It hasbeen reported that cell cycle regulation was the most widely affected pathway in HCC,and correlated best with the progression of cancer. In this paper, a cell cycle relatedHCC and cirrhosis classifier is developed based on K-TSP algorithm. This classifiershowed robust classification performances among datasets from different platforms anddifferent laboratories. On the other hand, the pathologic atypia can be indistinguishablefor pathologists and one cannot be certain that the sample did indeed come from thelesion in biopsy when lesion is so small in early hepatocellular carcinoma. To tackle theclinical difficulties in diagnosis of early hepatocellular carcinoma, we extract effectiveclassifier to identify precancerous lesions and early hepatocellular carcinoma based onthe relative gene expressions of cirrhosis tissue adjacent to hepatocellular carcinoma. Inthis paper, the algorithm used for extract biomarkers focus on the relative geneexpression values in individual sample, which is characterized by robust to platforms and different preprocess methods. The two classifiers also can be used for variousdatasets, combining then could serve as an effective early diagnosis aid.