Search

MicroRNAs expression by real-time qPCR. Figure S1. Principal component analysis of microRNAs expression and hierarchical clustering analysis in the exploratory cohort (A, C) and in the replication cohort (B, D). Figure S2. Pathways analysis of significant differentially expressed miRs in monocytes (A) and CD4+ T lymphocytes (B). Figure S3. Flow chart for the selection of published study for the comparison analysis. Figure S4. Cluster analysis of differentially expressed miRs in published studies from literature. (DOCX 2104 kb)

PCoA analysis of multiple tissue groups, related to Figs. 1, 2 and merging conditions section. Scatterplots of the first and second principal components from principal component analysis on all major tissue groups colored by sampled region. The grouping in these plots led us to either merge regions into a single group or to keep them separate. The final tissue set used for further analysis is summarized in Table 1. (PDF 73 kb)

Heatmap of the 15 most variable genes in the GTEx heart samples post filtering, related to Figs. 1 and 3. Heatmap of the 15 most variable genes in the GTEx heart samples. Left, top 15 genes were chosen in an unsupervised manner using the normalized gene expression after a stringent filtering in a tissue-agnostic manner. Right, the 15 most variable genes were chosen in an unsupervised manner using the normalized gene expression after tissue-specific filtering. (PDF 277 kb)

PCoA analysis of multiple tissues on Y-chromosomal genes can highlight poor sex annotation, related to Fig. 1 and misannotation section. Scatterplots of the first and second principal components from principal component analysis on all major tissue regions. We plotted data from 13 tissue regions from the GTEx consortium, coloring the annotated sex of each sample. Enlarged is sample GTEX-11ILO that clusters with male samples in every tissue despite being annotated as being from a female; we later learned that this research subject was genetically male. (PDF 240 kb)

Similarity between cell lines and their tissues of origin based on gene expression. (A) Principal component analysis (PCA) was performed to evaluate possible batch effects in the gene expression data. Samples are labeled based on the year the sample was analyzed by the GTEx project, and the plots show the sample separation for the first 7 PCs. (B) Number of genes expressed in each group (LCL, whole blood, fibroblast, skin). Genes were separated into biological classes using the definitions from GENCODE release 19 (GRCh37.p13). (C) PCA of paired samples between the two tissues and cell lines (total of 89 subjects with all four samples) based on the normalized expression of all genes. The primary axis separates samples by tissue; the secondary axis separates primary tissue from cell lines. (D) To access whether the PCA results were dependent on the 89 samples chosen because they were present in all four groups, we repeated the analysis 100 times using 89 randomly selected samples in each group. The left panel shows the projection of the first 2 PCs for one random analysis, and right panel shows the distribution of PC1 and PC2 for each of the 100 analyses. (PDF 267Â kb)

Count distributions pre- and post- normalization, related to Figs. 1 and 4. Density plots of gene count distributions. Left to right: log2 raw expression distribution of samples pre-normalization; count distribution for each sample normalized in a tissue-aware manner. Colors represent different tissues. (PDF 7035 kb)

Supplementary Material for Tissue-aware RNA-Seq processing and normalization for heterogeneous and sparse data. (DOCX 37 kb)

Differential expression analysis. (A) Volcano plots of the differential expression analysis using voom on paired samples between the indicated groups. The lines indicate a log2 fold change of −2 or 2. (B) Percentage of genes called differentially expressed (DE) varying the log2 fold change at a FDR

Gene expression variability. (A) Density plot of the gene expression standard deviation (SD) within each cell line/tissue group. (B) F-test was performed to evaluate the differences in gene expression variance between the indicated groups. The histograms show the ratio of variances at log scale for all the genes, and the red line indicates similar gene expression variance between the two indicated groups. The bar plots show the percentage of genes with significant differences in variance (FDRÂ

Reconstruction and robustness of gene regulatory networks. (A) A cartoon of how the networks were generated. We used PANDA, a message-passing network inference algorithm that integrates multiple types of genomic data and infers the network of interactions between TFs and their target genes. PANDA uses a prior regulatory network inferred by mapping TF binding sites to the genome (motif data), integrates protein-protein interaction data and group-specific gene expression data to iteratively refine and deduce a final regulatory network. We generated one PANDA network for each group: LCL, whole blood, fibroblasts, and skin. The illustrations represent an example subnetwork with 5 TFs and 50 of its target genes. The strength of the inferred regulatory relationship is indicated by the edge thickness. Next, we did multiple random selections of 40 paired samples, and generated 100 networks for each group: LCL, blood, fibroblast, and skin. (B) Density plot of the standard deviation of the edge weights across the 100 bootstrapped networks in each group: LCL, blood, fibroblast, and skin. (C) Scatter plot of the average edge weights obtained from the bootstrapped networks and the edge weights from the network obtained using all the samples. (D) Scatter plot of the TF out-degree differences between the indicated cell line and tissue for the bootstrapped networks versus the network obtained using all the samples. (E) Scatter plot of the gene in-degree differences between the indicated cell line and tissue for the bootstrapped networks versus the network obtained using all the samples. (PDF 1025Â kb)

Robustness analysis with respect to other epigenetic biomarkers of aging in the WHI. A-C Results for the Horvath method when 47 out of 353 CpGs were removed from the epigenetic clock (because they are in the vicinity of a SNP). Since none of the remaining clock CpGs are near a SNP, the resulting age acceleration is not trivially related to race/ethnicity. A DNA methylation age versus chronological age. B Ethnicity versus age acceleration (defined as residual resulting from regressing DNAm age on chronological age). C Intrinsic epigenetic age acceleration versus ethnicity. D, E Alternative epigenetic biomarker of aging based on 589 age-related CpGs from Teschendorff [13]. The biomarker was defined using the following steps. First, the DNA methylation levels of each CpGs were standardized (to mean zero and variance 1). Second, a weighted average was formed by multiplying each CpG by the T test statistic from the chronological age relationship based on the table from the original reference. Third, the weighted average was regressed on chronological age to arrive at a residual. The resulting residual is referred to as extrinsic measure of age acceleration since it was not adjusted for blood cell counts. Fourth, the resulting measure was regressed on estimated blood cell counts (analogous to those used for IEAA) in order to arrive an intrinsic measure of age acceleration. F, G Epigenetic measures of age acceleration using the Hannum method 71 CpGs [19]. D, F Results for intrinsic measures, i.e. measures of age acceleration that adjust both for blood cell counts and chronological age. E, G reports extrinsic measures, i.e. no adjustment for imputed blood cell counts. Each bar plot depicts 1 standard error and reports the results from a Kruskalâ Wallis test. (PDF 55 kb)

Educational level versus age acceleration in the WHI. Each row relates educational level (x-axis) to three respective measures of epigenetic age acceleration: (A-D) Age Accel., (E-H) IEAA, and (I-L) EEAA. The columns correspond to different groups of women from the WHI. The first, second, third, and fourth columns report findings for (A, E, I) all women, (B, F, J) Caucasians, (C, G, K) African Americans, and (D, H, L) Hispanics, respectively. Each bar plot reports the mean values, 1 standard error, and the p value from a non-parametric group comparison test (Kruskalâ Wallis). Education was assessed using the form â Demographics and Study Membership.â We find that education predicts future EEAA. (PDF 6 kb)

Epigenetic age acceleration in Hispanics versus country of residence in the WHI. Each column corresponds to different measure of age acceleration: (A, D) age acceleration residual, (B, E) IEAA (C, F) EEAA. (A-C, first row) results for “country of birth” (x-axis). (D-F, second row) results for “country of residence” at age 35 years, which was defined by combining two variables country of birth and “living in the US at age 35.” The left-most bar corresponds to Hispanic women who were born outside the US and lived outside the US at age 35 years, the middle bar corresponds to Hispanic women who were born outside the US but lived already in the US at the age of 35 years; the right-most bar reports results for women who were born in the US and lived in the US at age 35 years. Incidentally, all of these postmenopausal Hispanic women lived in the US at the age of the blood draw. As a caveat, we mention the relatively small group sizes (small gray numbers underneath the bars). (PDF 3 kb)

Lymphoblastoid cell lines from Han Chinese, Caucasians, and African Americans. A Gray line corresponds to a natural spline regression through Caucasian samples. Age acceleration was defined as residual with respect to this line. B Marginally significant evidence that African Americanâ s are younger than other ethnic groups. (PDF 33 kb)

Demographic and physiologic characteristics of women from the WHI. Case-control status refers to CHD. Two designs were used to select samples: case/control and case-cohort. (DOC 48 kb)

Type II diabetes status versus age acceleration in the Bogalusa study. Each row relates type II diabetes status (x-axis) to three respective measures of epigenetic age acceleration: (A-C) Age Accel., (D-F) IEAA, and (G-I) EEAA. The columns correspond to different groups. The first, second, and third columns report findings for (A, D, G) all participants, (B, E, H) Caucasians, (C, F, I) African Americans, respectively. Each bar plot reports the mean values, 1 standard error, and the p value from a non-parametric group comparison test (Kruskalâ Wallis). Type 2 diabetes status was defined as fasting glucoseâ >â =126Â mg/dl or taking diabetes medication. (PDF 3 kb)

Human populations have faced a large number of environmental challenges during their evolutionary history and present today a wide range of habitats and mode of subsistence. However, the extent of genetic adaptation and epigenetic responses to such environmental variation remains controversial. We first explored the power of several statistics to detect hard selective sweeps in the context of whole-genome sequencing data, and evaluated their robustness to demography and other selection modes. Using data from the 1,000 Genomes Project and Complete Genomics, we showed that hard sweeps targeting low-frequency standing variation have played a moderate, albeit significant, role in recent human evolution. The signals of selection detected were moreover enriched in functional variants detected by genome-wide association studies. We then evaluated the relative impacts of genetic and environmental factors on human epigenomic diversity. To do so, we generated genome-wide genetic and DNA methylation profiles for Central African populations differing in their current habitat or in their historical lifestyle and genetic background. We found that both factors have similar critical impacts on the shaping of the global methylome, but the biological functions affected and the mechanisms underlying DNA methylation variation strongly differ. More generally, methylation variation shows strong associations with nearby genetic variants that, moreover, are enriched in signals of natural selection. Together, this work provides new insight into the contribution of genetic adaptation and epigenetic responses to the adaptation of humans to environmental changes over different time scales., Les populations humaines ont été confrontées à de nombreux changements environnementaux au cours de leur histoire et présentent aujourd’hui une grande diversité d’habitats et de modes de subsistance. Cependant, l’ampleur de l’adaptation génétique et des réponses épigénétiques à ces changements est débattue. Nous avons d’abord étudié la puissance de diverses statistiques pour détecter les balayages sélectifs dans le contexte des données de séquençage à haut débit, et évalué leur robustesse à différents facteurs confondants. En utilisant des jeux de données de séquençage, nous montrons que les balayages sélectifs ont eu un impact modéré mais non négligeable dans l’évolution récente du génome humain. Les régions sous sélection sont enrichies en mutations associées à des variations phénotypiques. Nous avons ensuite évalué l’impact respectif des facteurs génétiques et environnementaux sur la diversité épigénétique humaine. Pour cela, nous avons obtenu les génotypes et les profiles de méthylation de l’ADN de populations d’Afrique Centrale présentant des différences récentes d’habitat ou historiques de modes de vie et de profil génétique. Nous montrons que les deux facteurs ont un effet similaire sur le méthylome mais diffèrent par les fonctions biologiques affectées et les mécanismes expliquant les variations observées. Plus généralement, les variations de méthylation sont fortement associées à des mutations génétiques qui sont enrichies en signaux de sélection positive. En conclusion, ce travail apporte un aperçu de la contribution des mutations génétiques et des réponses épigénétiques à l’adaptation humaine aux changements environnementaux sur plusieurs échelles de temps.