Figure 1.
Performance of seriation on simulated SAGE data.
(A). Seriation results of the three rounds of simulations with increasing amounts of noise from round 1 (34 tags) to round 3 (384 tags). The dark red squares along the diagonal indicate tags with the expression patterns 1–3 that were grouped together by seriation. (B). Seriation of 10 expression profiles with limited amount of noise. The dark red squares along the diagonal indicate tags in each expression profile that were grouped together. The numbers indicate expression patterns from Figure S1 that were grouped into each contig. Note that two contigs in the middle (5 and 1) appear more similar to each other than any other contig pair indicating similarity of the corresponding expression patterns.
Table 1.
Effect of the amount of noise in SAGE data on the performance of seriation and PoissonC.
Table 2.
Comparative performance of seriation and PoissonC on a simulated SAGE data set with 10 expression patterns.
Figure 2.
Seriation of genes expressed in mouse retinal SAGE libraries.
SAGE data from Blackshaw et al. [20] were subjected to seriation analysis as described in the text. The resulting reordered correlation matrix containing correlation coefficients for each tag pair computed to measure the similarity of their retinal expression profiles is color-coded red to blue to represent decreasing correlation values. Ten contigs, including two supercontigs, recognizable as the squares of high (red) correlation values along the diagonal, are evident from the color-coded correlation matrix. The Figure on the right provides a zoomed-in view of the contigs.
Figure 3.
Analysis of seriation contigs of genes expressed in mouse retinal SAGE libraries.
(A). Comparison of seriation contigs to the original clusters from Blackshaw et al. [20]. Seriation contigs are color-coded and plotted on the x-axis of the 3D graph. The peaks on the z-axis represent the percent cluster members (y-axis) present in the particular contig. Most seriation contigs are composed of one or several predominant clusters (also see Table 3). (B). Expression profiles of genes in seriation contigs. The relative expression levels from 0% to 100% are plotted on the y-axis for each contig while the retinal libraries derived from developmental stages E12.5, E14.5, E16.5, E18.5, P0.5, P2.5, P4.5, P6.5, P10, and adult are on the x-axis. The ordering of contigs is temporal such that genes expressed in earlier developmental stages tend to be in the first contigs, while genes expressed in later stages are in later contigs. This partitioning is particularly evident from the expression patterns of genes in the supercontigs.
Table 3.
Comparison of seriation and PoissonC analyses of genes expressed in retinal SAGE libraries.
Figure 4.
Seriation of transcription factors expressed in Mouse Atlas pancreatic libraries.
SAGE data for transcription factors expressed in the pancreatic libraries from the Mouse Atlas project were subjected to seriation analysis as described in the text. The reordered correlation matrix containing correlation coefficients for each tag pair computed to measure the similarity of their pancreatic expression profiles is color-coded red to blue to represent decreasing correlation values. (A). 5 contigs recognizable as red squares along the diagonal are evident. (B). Expression profiles of transcription factors in contigs in (A). The relative expression levels from 0% to 100% are plotted on the y-axis for each contig while the pancreatic libraries derived from stages TS17, TS19, TS20, TS21, TS22, and P70 are on the x-axis.
Table 4.
Summary of seriation analysis of transcription factors expressed in pancreas.