204 Chapter 8 (p<0.05 as calculated by Wald test)5 and are referred to as ‘differentially acetylated regions’ (Table S2A). Supervised hierarchical clustering was performed with quantile normalised (limma::normalizeQiantiles() function in R), log2 transformed and median centred read counts per common region. To avoid the log2 transformation of zero values, one read was added to each region. Gene annotated to differentially acetylated regions and pathway analysis: Region to gene annotation was performed in silico using a conservative window of +/-5kb from the transcription start site, Table S2B and 2C. ToppFun and STRING were used for gene list enrichment analysis and candidate gene prioritisation based on functional annotations and protein interaction networks.6,7 For ToppFun, the list of hyper- and hypoacetylated genes was tested using probability density function p-value calculation method, FDR B&H correction, p-value cut-off of 0.05, and gene limit of 1-2,000 genes per term (Table S2D). Since the prebuild gene/protein networks integrated into ToppFun were not created using the same criteria and the 126 annotated number of genes varies significantly, we also reported genes belonging to known disease pathways even below the p-value threshold (where indicated). For protein network interaction visualisation STRING v10.0 was used with a minimum required interaction score at the highest confidence setting for all differentially acetylated peaks. Enrichment of TFBMs in differentially acetylated regions: A total of 3,396 cardiac DNAse hypersensitivity sites (DHS) were obtained from the ENCODE database (Heart_OC, Primary frozen heart tissue from NICHD donor ID:1104, Male, Caucasian, 35 years old)8 overlapping with all differentially acetylated regions in the PLN-R14del vs. control group (both hypo- and hyperacetylated) were used for this analysis. The genomic sequence of DHS was repeat masked and the enrichment of TFBM was calculated against the shuffled sequences using the Analysis Motif of Enrichment (AME tool) of the MEME Suite with the following settings: motif database: human (HOCOMOCO v9), background model sequence set to 0.29182,0.20818,0.20818,0.29 182, pseudo count added to a motif column: 0.25, Wilcoxon rank-sum test (quick), p<0.05, Table S3A), number of multiple tests for Bonferroni correction: #Motifs× #PartitionsTested = 426×1 = 4269. The functional annotation of the enriched TFs was performed as described above (Table 141 S3B). Protein network interaction was performed as explained above using the high confidence interaction score. PLN-specificity analysis of differentially acetylated regions: Sequencing reads from each ChIPseq sample (PLN-R14del (n=6), control (n=4), ischemic (n=4), and sarcomeric (n=6) groups) were compared to the common differentially acetylated region list to set the H3K27ac occupancy for every region-sample pair. Raw read counts were quantile normalised (limma::normalizeQiantiles() function in R), log2 transformed and median centred (to avoid log2 transformation of zero values, one read was added to each region). The median value from each sample group was used to construct an n x k table where n = 4 (one value per 150 each sample type) and k represent the number of differentially acetylated regions. The
RkJQdWJsaXNoZXIy MTk4NDMw