van Dijk and the Adessium Foundation. For most of these cell-types multiple biological replicates have been assayed. We quantile normalized, log2 transformed and then centered the expression levels for every individual gene to a imply of zero and a standard deviation of one. We observed that 42 of these 44 genes show significantly higher expression (Student’s T-test P < 0.001), as compared to the other 13 cell types. NS: non significant.(TIF) pgen.1005223.s002.tif (2.3M) GUID:?B6A45B31-BF41-4978-8E97-4D38810DFFAA S3 Fig: Relationship between neutrophil percentage, age and gender. We correlated the actual neutrophil percentage (top) and the inferred neutrophil percentage (bottom) with age in the EGCUT dataset (n = 825) and observed that there is a low, but significant correlation between age and neutrophil percentage. However, neutrophil percentage is not significantly associated with gender.(TIF) pgen.1005223.s003.tif (970K) GUID:?5DD03A0A-D1B8-40BF-9492-8A8EEED618D4 S4 Fig: Stability of Cloprostenol (sodium salt) neutrophil percentage prediction. We tested the stability of Cloprostenol (sodium salt) our neutrophil percentage prediction in the EGCUT dataset (n = 825). From your list of 100 probes showing highest correlation with neutrophil percentage, we randomly selected a number of probes (increments of 5 probes, 1000 permutations per increment) and repeated the neutrophil percentage prediction. When including > 10 probes, the neutrophil prediction displays stable correlation with the actual neutrophil percentage (Spearman R ~0.75) and near ideal correlation with the predicted neutrophil percentage used in the meta-analysis (Spearman R ~0.99). Error bars denote standard deviation. Red collection denotes the number of gene expression probes the different cohorts in this study used to estimate neutrophil percentage.(TIF) pgen.1005223.s004.tif (471K) GUID:?A599001C-431E-4F4B-ACE2-AECA94DD3235 S5 Fig: gene expression levels, age and gender. We correlated the actual gene expression levels with age in the EGCUT Rabbit polyclonal to ANGPTL4 dataset (n = 825, normalized using log2 transformed and quantile normalization, and gene expression levels corrected for 40 principal components) and observed that there is a low, but significant correlation between age and gene expression in the log2 transformed and quantile normalized data (top), which becomes insignificant when correcting the gene expression data for 40 principal components (which was used to determine the neutrophil conversation effect; bottom). However, gene expression levels are not significantly associated with gender.(TIF) pgen.1005223.s007.tif (979K) GUID:?0764B8CA-08B4-4095-81F1-5F1C240C37D7 S8 Fig: Effect of strong estimation of standard errors. The conversation model we used does not take heteroscedasticity into account. Therefore, we decided standard errors using the ‘sandwich’ package in R, which allows for the estimation of strong standard errors. We observed strong correlation between standard errors, Z-scores and p-values by our model and a model that applies strong estimation of standard errors in the EGCUT (top) and Fehrmann datasets (bottom).(TIF) pgen.1005223.s008.tif (1.0M) GUID:?EB79A7B5-6051-4F66-BCD1-EEABFE16077A S9 Fig: Principal components on gene expression data. Principal component 1 (PC1) and principal component 2 per study. Samples with a correlation < 0.9 with PC1 (red) were excluded from analysis.(TIF) pgen.1005223.s009.tif (1.1M) GUID:?9AB25CD7-C0C4-41CF-94F5-4ECFB598F5D8 S10 Fig: Neutrophil percentage and principal component correction. The gene expression data that was utilized for the conversation meta-analysis was corrected for up to 40 principal components. In order to retain genetic variance in the gene expression data, components that showed a significant correlation with genotypes were not removed. In the EGCUT dataset (n = 825), many of these components also strongly correlate with neutrophil percentage (top) and inferred neutrophil percentage (bottom). The majority of the variance in gene expression explained Cloprostenol (sodium salt) by these components (right) was however removed from this dataset.(TIF) pgen.1005223.s010.tif (1.3M) GUID:?E1079246-F880-4B58-9067-2CF510EC76BE S1 Table: List of 58 Illumina HT12v3 probes utilized for calculating the estimated neutrophil percentage principal component score and their correlation with neutrophil percentage in the EGCUT dataset (n = 825). (XLSX) pgen.1005223.s011.xlsx (41K) GUID:?457E86F0-6936-4C6D-988F-0612D90B9FE9 S2 Table: Summary statistics for the interaction analysis. (XLSX) pgen.1005223.s012.xlsx (8.4K) GUID:?2FAE8FE1-FFE0-4C95-9663-ECE417021D4F S3 Table: Results of the conversation analysis. (XLSX) pgen.1005223.s013.xlsx (1.7M) GUID:?B37CE8D7-BB68-4F38-BEEC-6275420FA590 S4 Table: Summary statistics showing the effect size (correlation coefficient) in each of the tested replication datasets. (XLSX) pgen.1005223.s014.xlsx (1.8M) GUID:?2EBE5F4D-6077-428A-B707-8F0281864DB9 S5 Table: Results of the neutrophil mediated.
van Dijk and the Adessium Foundation