Abstract
AIM: An in silico pathway analysis was performed in an attempt to identify new biomarkers for cervical carcinoma. METHODS: Three publicly available Affymetrix gene expression data sets (GSE5787, GSE7803, GSE9750) were retrieved, vouching for a total 9 cervical cancer cell lines, 39 normal cervical samples, 7 CIN3 samples and 111 cervical cancer samples. An Agilent data set (GSE7410; 5 normal cervical samples, 35 samples from invasive cervical cancer) was selected as a validation set. Predication analysis of microarrays was performed in the Affymetrix sets to identify cervical cancer biomarkers. We compared the lists of differentially expressed genes between normal and CIN3 samples on the one hand (n=1923) and between CIN3 and invasive cancer samples on the other hand (n=628). RESULTS: Seven probe sets were identified that were significantly overexpressed (at least 2 fold increase expression level, and false discovery rate <5%) in both CIN3 samples respective to normal samples and in cancer samples respective to CIN3 samples. From these, five probes sets could be validated in the Agilent data set (P<0.001) comparing the normal with the invasive cancer samples, corresponding to the genes DTL, HMGB3, KIF2C, NEK2 and RFC4. These genes were additionally overexpressed in cervical cancer cell lines respective to the cancer samples. The literature on these markers was reviewed. CONCLUSION: Novel biomarkers in combination with primary human papilloma virus (HPV) testing may allow complete cervical screening by objective, non-morphological molecular methods, which may be particularly important in developing countries.