Karlijn Hummelink

Chapter 5 178 set had 100 samples with protein expression data from a panel of 1,305 proteins. (Protein expression measurements were generated by SomaLogic, Boulder, CO, USA.) Specific protein sets were created as the intersection of the list of the panel targets and results of queries for biological functions from GeneOntology, using AmiGO2 tools (http://amigo.geneontology.org/amigo) and UniProt databases (https://www. uniprot.org/). The PSEA method associated test classification with these biological functions via a rank-based correlation of the measured protein expressions with the test classifications of the reference samples23. The mass spectral features associated with biological processes (in particular Immune Response Type 2) were determined using Spearman correlation of the measured protein expressions with the mass spectral features23 using the 49-sample reference set only. While the implementation closely follows the GSEA approach, we employed an extension of the standard method that increases the statistical power to detect associations between phenotype (test classification subgroup) and biological process24. The PSEA was carried out using a C# implementation and Matlab (MathWorks, Natick, MA, USA). PSEA P-values were defined as described by Subramanian and colleagues20. False discovery rates (FDRs) for the PSEA calculations were assessed using the method of Benjamini and Hochberg25. Other Statistical analysis All analyses, except the PSEA, were carried out using SAS9.3 (SAS Institute, Cary, NC, USA) or PRISM (GraphPad, La Jolla, CA, USA). Survival/progression-free survival plots and medians were generated using the Kaplan–Meier method. Association between test classification and categorical or continuous variables was assessed using Fisher’s exact test and Mann-Whitney test, respectively. All P-values are two-sided.

RkJQdWJsaXNoZXIy MTk4NDMw