TY - GEN
T1 - Can specific transcriptional regulators assemble a universal cancer signature?
AU - Roy, Janine
AU - Isik, Zerrin
AU - Pilarsky, Christian
AU - Schroeder, Michael
PY - 2013
Y1 - 2013
N2 - Recently, there is a lot of interest in using biomarker signatures derived from gene expression data to predict cancer progression. We assembled signatures of 25 published datasets covering 13 types of cancers. How do these signatures compare with each other? On one hand signatures answering the same biological question should overlap, whereas signatures predicting different cancer types should differ. On the other hand, there could also be a Universal Cancer Signature that is predictive independently of the cancer type. Initially, we generate signatures for all datasets using classical approaches such as t-test and fold change and then, we explore signatures resulting from a network-based method, that applies the random surfer model of Google's PageRank algorithm. We show that the signatures as published by the authors and the signatures generated with classical methods do not overlap - not even for the same cancer type - whereas the network-based signatures strongly overlap. Selecting 10 out of 37 universal cancer genes gives the optimal prediction for all cancers thus taking a first step towards a Universal Cancer Signature. We furthermore analyze and discuss the involved genes in terms of the Hallmarks of cancer and in particular single out SP1, JUN/FOS and NFKB1 and examine their specific role in cancer progression. © 2013 AIP Publishing LLC.
AB - Recently, there is a lot of interest in using biomarker signatures derived from gene expression data to predict cancer progression. We assembled signatures of 25 published datasets covering 13 types of cancers. How do these signatures compare with each other? On one hand signatures answering the same biological question should overlap, whereas signatures predicting different cancer types should differ. On the other hand, there could also be a Universal Cancer Signature that is predictive independently of the cancer type. Initially, we generate signatures for all datasets using classical approaches such as t-test and fold change and then, we explore signatures resulting from a network-based method, that applies the random surfer model of Google's PageRank algorithm. We show that the signatures as published by the authors and the signatures generated with classical methods do not overlap - not even for the same cancer type - whereas the network-based signatures strongly overlap. Selecting 10 out of 37 universal cancer genes gives the optimal prediction for all cancers thus taking a first step towards a Universal Cancer Signature. We furthermore analyze and discuss the involved genes in terms of the Hallmarks of cancer and in particular single out SP1, JUN/FOS and NFKB1 and examine their specific role in cancer progression. © 2013 AIP Publishing LLC.
UR - http://www.scopus.com/inward/record.url?scp=84887864557&partnerID=8YFLogxK
U2 - 10.1063/1.4825005
DO - 10.1063/1.4825005
M3 - Conference contribution
T3 - AIP Conference Proceedings
SP - 135
EP - 144
BT - 2013 International Symposium on Computational Models for Life Sciences
T2 - 2013 International Symposium on Computational Models for Life Sciences, CMLS 2013
Y2 - 27 November 2013 through 29 November 2013
ER -