This table summarizes the relevant GSEA results obtained for the different comparisons (19 AITLs versus 16 PTCLs-u; 2 AITL cell suspensions versus 17 AITL tissue samples; 6 CD30⁺ versus 9 CD30⁻ PTCLs-u; 19 AITLs versus 9 CD30⁻ PTCLs-u). Shown are the enriched gene sets with corresponding size, the maximal P value, and the enriched class of samples. A weighted enrichment score (ES) is calculated by walking down the ranked gene list, increasing a running-sum statistic for every gene set encountered and decreasing the statistic for every gene not in the gene set. A P value is calculated for an ES by using a sample-based permutation test procedure that preserves the complex correlation structure of the gene expression data. Specifically, we permuted the samples labels (eg, PTCL-u and AITL) and recomputed the ES of the gene set for the permuted data 1000 times, which generated a null distribution for the ES. The P value of the observed ES was then calculated based on this null distribution. Other advanced parameters used were as follows: “meandiv” normalization of the data; “timestamp” seed for permutation; and “equalize and balance” randomization mode. The size of a given gene set for a given comparison represents the number of genes out of the total genes in the gene set that were present in the data set following pre-GSEA filtering. After reducing the data set to account for redundant probe sets for a single gene identifier, we obtained the following data sets for each comparison: 13 178 genes for AITL versus PTCL-u; 13 186 genes for AITL versus CD30⁻PTCL-u; and 13 244 genes for CD30⁺ PTCL-u versus CD30⁻ PTCL-u. Since slightly different data sets were used for each comparison (depending on a minimal geometric mean of the 2 groups of samples being compared; see “Patients, materials, and methods”), the number of genes in the table did not necessarily correspond to the number of genes in a gene set that were represented in a data set used in the comparison. For the AITL versus PTCL-u and CD30⁺ versus CD30⁻ PTCL-u comparisons, SNR was used to construct the ordered data set. For comparison of the AITL tissues and sorted cells, genes were ordered according to fold change of expression.

A meaningful P value cannot be derived from the permutation tests because the number of samples in one class (cell suspensions) was too small (n = 2).