We thank Drs. Liu and Ye for their interest in our study. We appreciate their concern and would like to reiterate that the strength of our risk score, and its implication in the prognostication of hepatocellular carcinoma, is the result of unbiased systematic selection of a limited number of genes from a large number of genes in previously established prognostic signatures.[2, 3] As Drs. Liu and Ye correctly pointed out, weak association of risk score with prognosis in data from Hoshida et al's study might have been the result of a difference in platform and sample sources. Our risk score was developed with data generated from frozen tissues. However, data from GSE10141 were generated by applying complementary DNA-mediated annealing, selection, extension, and ligation technology with the use of fragmented RNAs from formalin-fixed paraffin-embedded tissues. Moreover, the number of patients in their test set is smaller than our cohorts. Clinical data from one of the validation data sets (GSE14520) used in our study were recently updated and available as extra supplement data from the Gene Expression Omnibus. This data set would be better for the evaluation of different prognostication methods. However, we agree with them that new prognostication methods should be validated in a prospective study.
EUN SUNG PARK, PH.D.1
JU-SEOG LEE, PH.D.2
1Institute for Medical Convergence, Yonsei University College of Medicine, Seoul, Korea
2Department of Systems Biology, The University of Texas MD Anderson Cancer Center, Houston, TX