Comparative transcriptome map of human ES and EB on chromosome 17. The transcriptome map shows the coexpression index value of genes displayed according to their positions along the chromosome (horizontal axis). Each dot represents a gene expressed in ES (blue) or EB (red) cells. The yellow line represents the threshold coexpression index (0.3). The coexpression index was calculated as the average Pearson's correlation coefficient between the expression level of a given gene with that of every neighboring gene (10 up- and 10 downstream). The chromosomal position on the horizontal axis is illustrated by both coordinates (mega base pair) and the ideogram with markers of cytogenetic bands [20]. The statistical significance of the coexpression pattern was confirmed by Monte-Carlo simulation (see Methods). A. Transcriptome map constructed with the real genome data, where multiple peaks of coexpression above the threshold are visible and appear as non-random. The co-expressed genes include, from left to right, ORIA1, SOX15, ALDH3A2, PIPOX, KRTHA8, STAT3, COX11, MIRPL38, and PCYT2 (indicated by arrows). B. Transcriptome map by randomized genome data, where peaks of coexpression are random and few are above the threshold.

Coexpression chromosomal domains of SOX15 (A), PTPRCAP (B), and NGFRAP1 (C), which represent three different coexpression patterns in ES and EB. Left, principal component analysis map. Each dot represents a gene of a domain expressed in ES (blue dots) or EB (brown dots). The wire mesh was constructed based on a standard error of 2. Genes with correlated expression profiles tend to cluster tightly together into a small-size ellipsoid by the wire mesh, while genes with less similar expression profiles form a looser cluster with larger size of ellipsoid. The SOX15 domain is shown to have a higher coexpression level in ES than EB, PTPRCAP domain has a higher coexpression in EB than ES, while NGFRAP1 domain shows a non-differential coexpression pattern between ES and EB. Middle, transcriptome map of the chromosome domain (genomic locations are in mega base pairs). Each gene on the heatmap is indicated. In the SOX15 domain, most genes have higher coexpression index values in ES than in EB. In the PTPRCAP domain, most genes have higher coexpression index values in EB than ES. In the NGFRAP1 domain, most genes have similar coexpression index values between ES and EB. Right, heatmap of gene expression and unsupervised hierarchical cluster analysis. Each column in the heatmap corresponds to an ES or EB sample. Each row represents the relative level of expression of a single gene. The genes are arranged from top to bottom in the same order as they appear on the chromosome from 5' to 3' shown on the transcriptome map. Each gene is denoted by its symbol, followed by the fold-change and ANOVA P value (in brackets) indicating the level of differential expression of the gene between ES and EB. Red, high expression relative to the mean expression; green, low expressions relative to the mean expression; black, no significant change in expression level between the sample and mean.

The distributions of functional semantic similarity scores derived from coexpression chromosome domains in ESC and EB, and from a randomized gene set. The green line on the plot represents the distribution from the random data, the blue and red lines are those from ESC and EB, respectively. The X-axis indicates functional semantic similarity scores, and the Y-axis indicates the accumulative probability of coexpressed genes in domains. For the random gene set, the cumulative probability reached to 1 when the functional similarity was up to 8. That means all the random gene pairs have the functional similarity 8 or below. For coexpression domain genes of ESC and EB, the cumulated probability did not reaches 1 (i.e. 100% of gene pairs) until the functional similarity was over 35, indicating higher functional similarities from more coexpressed genes. The distributions from ESC and EB domain gene sets were similar to each other but significantly different from that from the random gene set (p < 10E-10 by the Kolmogorov-Smirnov test).

Genomic overview of coexpression chromosomal domains in human ES and EB cells. The domains were identified by the genes which have significant correlation in the expression profile with adjacent genes (10 upstream and 10 downstream) along the chromosome, at a coexpression index threshold of 0.3. Blue bars on the chromosome represent coexpression domains in ES, while red bars represent coexpression domains in EB.

Coexpression chromosomal domains of SOX15 (Fig 3), PTPRCAP (Fig 4), and NGFRAP1 (Fig 5), which represent three different coexpression patterns in ES and EB. Left, principal component analysis map. Each dot represents a gene of a domain expressed in ES (blue dots) or EB (brown dots). The wire mesh was constructed based on a standard error of 2. Genes with correlated expression profiles tend to cluster tightly together into a small-size ellipsoid by the wire mesh, while genes with less similar expression profiles form a looser cluster with larger size of ellipsoid. The SOX15 domain is shown to have a higher coexpression level in ES than EB, PTPRCAP domain has a higher coexpression in EB than ES, while NGFRAP1 domain shows a non-differential coexpression pattern between ES and EB. Middle, transcriptome map of the chromosome domain (genomic locations are in mega base pairs). Each gene on the heatmap is indicated. In the SOX15 domain, most genes have higher coexpression index values in ES than in EB. In the PTPRCAP domain, most genes have higher coexpression index values in EB than ES. In the NGFRAP1 domain, most genes have similar coexpression index values between ES and EB. Right, heatmap of gene expression and unsupervised hierarchical cluster analysis. Each column in the heatmap corresponds to an ES or EB sample. Each row represents the relative level of expression of a single gene. The genes are arranged from top to bottom in the same order as they appear on the chromosome from 5' to 3' shown on the transcriptome map. Each gene is denoted by its symbol, followed by the fold-change and ANOVA P value (in brackets) indicating the level of differential expression of the gene between ES and EB. Red, high expression relative to the mean expression; green, low expressions relative to the mean expression; black, no significant change in expression level between the sample and mean.

Coexpression chromosomal domains of SOX15 (A), PTPRCAP (B), and NGFRAP1 (C), which represent three different coexpression patterns in ES and EB. Left, principal component analysis map. Each dot represents a gene of a domain expressed in ES (blue dots) or EB (brown dots). The wire mesh was constructed based on a standard error of 2. Genes with correlated expression profiles tend to cluster tightly together into a small-size ellipsoid by the wire mesh, while genes with less similar expression profiles form a looser cluster with larger size of ellipsoid. The SOX15 domain is shown to have a higher coexpression level in ES than EB, PTPRCAP domain has a higher coexpression in EB than ES, while NGFRAP1 domain shows a non-differential coexpression pattern between ES and EB. Middle, transcriptome map of the chromosome domain (genomic locations are in mega base pairs). Each gene on the heatmap is indicated. In the SOX15 domain, most genes have higher coexpression index values in ES than in EB. In the PTPRCAP domain, most genes have higher coexpression index values in EB than ES. In the NGFRAP1 domain, most genes have similar coexpression index values between ES and EB. Right, heatmap of gene expression and unsupervised hierarchical cluster analysis. Each column in the heatmap corresponds to an ES or EB sample. Each row represents the relative level of expression of a single gene. The genes are arranged from top to bottom in the same order as they appear on the chromosome from 5' to 3' shown on the transcriptome map. Each gene is denoted by its symbol, followed by the fold-change and ANOVA P value (in brackets) indicating the level of differential expression of the gene between ES and EB. Red, high expression relative to the mean expression; green, low expressions relative to the mean expression; black, no significant change in expression level between the sample and mean.