Translating bioinformatics in oncology: Guilty by profiling meta-analysis and identification of KIF18B and CDCA3 as novel driver genes in liver carcinogenesis

T Itzel; P Scholz; T Maass; M Krupp; JU Marquardt; S Strand; D Becker; F Staib; H Binder; XW Wang; SS Thorgeirsson; PR Galle; A Teufel

doi:10.1055/s-0032-1332071

Zeitschrift für Gastroenterologie, Table of Contents

Translating bioinformatics in oncology: Guilty by profiling meta-analysis and identification of KIF18B and CDCA3 as novel driver genes in liver carcinogenesis

T Itzel ¹, P Scholz ¹, T Maass ¹, M Krupp ¹, JU Marquardt ¹, S Strand ¹, D Becker ¹, F Staib ¹, H Binder ², XW Wang ³, SS Thorgeirsson ³, PR Galle ¹, A Teufel ¹

¹Johannes Gutenberg University, Department of Medicine I, Mainz, Germany
²Johannes Gutenberg University, 2Institute of Medical Biostatistics, Epidemiology and Informatics (IMBEI), Mainz, Germany
³National Institutes of Health, National Cancer Institute, Bethesda, MD, USA

Abstract

Gene expression profiling studies in cancer research have traditionally investigated relevant transcriptomics changes from single patient cohorts. However, co-regulated genes are not identified in such analyses but may theoretically be closely functionally linked (guilt by association, guilt by profiling). Although bioinformatics procedures for guilt by profiling/association analyzes were previously reported, this comprehensive approach has not been applied to large scale cancer biology yet.

Here, we analyzed the complete GSE2109 data repository of 2158 full cancer transcriptomes from 163 diverse cancer entities using Pearson's correlation coefficient for similarity of gene expression. Subsequently 428 highly co-regulated genes (>=0,8) were unsupervised clustered to obtain small co-regulated networks and further characterized by means of gene ontology and signalling pathway analysis.

One major subnetwork containing 61 closely co-regulated genes showed highly significant enrichment of biofunctions relevant to carcinogenesis. Within this subnetwork all genes except for KIF18B and CDCA3 had a previously confirmed tumor-biologic relevance. Therefore, we independently analyzed differenzial regulation of these two genes in multiple tumors and demonstrated a severe deregulation of both genes in breast, lung, ovary and kidney cancer proving our guilt by association hypothesis. Overexpression of KIF18B and CDCA3 in hepatoma cells and subsequent microarray analysis revealed a significant deregulation of central cell cycle regulatory genes as well as key check point kinases such as CyclinB1, CyclinB2, Cdk2, Cdk4. Consistently, FACS cell cycle analyses and proliferation assays confirmed the role of both genes during G2/M progression.

Finally, a prognostic significance for the identified KIF18B signature (p=0.03) and a clear trend for the CDCA3 signature (p=0.09) was demonstrated in two independent cohorts of >250 HCC patients as well as multiple other tumors. In summary, we present evidence for the usefulness of large scale guilt by profiling/association strategies in oncology. We identified two novel oncogenes, demonstrated their deregulation in multiple tumors and functionally characterized these novel oncogenes. Moreover, the robust prognostic importance of the downstream signatures for HCC and multiple other tumors indicates the clinical relevance of our findings.