Bioinformatic and Statistical Analysis

The raw data were processed with Illumina GenomeStudio V2011.1 (Genotyping Module v1.9.4). We used the reference panel with 96 normal samples in the discovery panel and with 816 normal samples in the replication panel to normalize raw data and generate log ratios. All genotyped samples were retained for this step. Single-nucleotide polymorphism (SNP), Log R Ratios (LRR) and B allele frequencies were calculated from the GenomeStudio software (Illumina Inc., San Diego, CA) with default settings for each panel separately. For subsequent analyses, only individuals were retained, where all 3 processed samples fulfilled the quality criteria of a standard deviation of the LRR less than 0.3 and standard deviation of B allele frequency less than 0.05. This lead to the analysis of 8 samples in the discovery step and 17 samples in the replication step.

Genomic aberrations in tumor samples were detected using OncoSNP v2.1.³² We analyzed each sample separately for autosomal chromosomes and with a maximum of 12 different tumor states. Phylogenies of the 2 HCC nodules and nontumorous control liver tissue for each patient were calculated using Medicc.³³ Only copy number regions supported by at least 40 SNPs were used for calculating the minimal event distances. From each phylogenetic tree, the distance from the control tissue to the branch point (X) and the length of the branches connecting the 2 tumors was extracted (Y=Y1 + Y2, Figure ​Figure11).

Illustration of the evolutionary distance measures. NT denotes the cirrhotic liver without tumor, that is, the “nontumoral tissue” of the patient. The 2 samples HCC nodules per patient are denoted as TU1 and TU2. Distances are calculated as Y, denoting the total distance between the 2 HCC nodules, and D, denoting the total evolutionary distance spread in the patient. The calculated phylogenetic tree defined the distance from NT to the branch point (X) and the distances of TU1 and TU2 to the branch point.

Joint analysis of clinical variables and measures of phylogenetic distance was performed using R (www.r-project.org).³⁴ Comparisons of distances between groups were performed using the Wilcoxon test and the Student t test. Predictors of survival after transplantation were performed using Cox regression.