Dear Editor, Hepatocellular carcinoma (HCC) is ranked as the most prevalent subgroup of liver malignancies in the world, representing about 90% of primary liver cancers. 1 To date, no widely accepted molecular biomarkers are available for survival stratification with HCC. 2,3 Recently, a novel algorithm was developed based on the relative orderings of mRNA expression patterns and had yielded excellent results. 4,5 Substantial breakthroughs have been founded in programmed death 1 (PD-1)/ programmed death 1 ligand (PD-L1) signaling pathway in various cancers. 6 Here, we used the TCGA cohort to develop a signature and other two databases to confirm the prognostic model based on pairwise PD-1/PD-L1 signaling pathway genes.As shown in the flowchart (Figure 1A), the mRNA expression matrixes and their clinical characteristics were retrieved from the TCGA, GEO, and the International Cancer Genome Consortium (ICGC-JP cohort) databases. The clinicopathological information of the three databases was displayed in Table S1. Finally, 146 potential genes were acquired from the KEGG and Reactome pathway databases (Table S2). We filtered out the candidate genes computed by median absolute deviation (MAD) > 0.5. We created all possible pairs of genes based on the shared 34 genes. Among gene pairs, two genes (a and b) formed a gene pair in alphabetical order. If the expression values of gene a > gene b in a particular sample, the gene pair produced a score of 1; if the situation is opposite, the score equals 0. The association of these gene pairs was evaluated with univariate and LASSO Cox regression analysis after 1000 iterations (Figure 1B-C), and finally, 12 gene pairs, included 13 genes, were identified (Table S3). The risk score was computed via the expression patterns of these pairwise genes weighted by the respective coefficient obtained from LASSO algorithm. On the basis of time-dependent ROC curve, the ideal cutoff of the risk score was calculated at 0.573 to stratify individuals into the high-andThis is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.