Fig 1.
On top, the first two principal components. The panel on the left illustrates the clear separation along the X-axis of the hematopoietic and non-hematopoietic material containing samples from both tissues and cell lines. The non-solid region includes blood, bone marrow, lymph nodes, tonsil, osteoclasts, spleen, sputum, thymus gland, bronchoalveolar lavage cells and derived cell lines. An elongated cluster including incompletely differentiated cells is also found. In the right panel, the Y-axis separates cell lines (top), neoplasias (middle) and non-neoplastic diseases (bottom), whereas normal tissues overlap with all three. The panel in the bottom shows the data for the first (X-axis) and third components (Y-axis). The hematopoietic axis (X-axis) allows detaching leukaemias from other blood neoplasias. Cell lines derived from brain tumours can be distinguished from their tissues of origin along this axis as well. The Y-axis detaches non-neoplastic central nervous system samples from tumoral ones, and separates myelomas from lymphomas.
Fig 2.
Clustering of biological groups with at least 20 observations.
a) Clustering of biological groups using average correlations. Only the 20,000 most variable probesets are accounted for. The groups are recoded by colours to display the largest clusters in the dendrogram. The colour labels on the right hand side specify groups of healthy, cancerous or diseased tissues or cell lines. Illustrative groups of tissues are also highlighted. b) Clustering of biological groups of solid tissues against the 1,000 most variable probesets. Many clusters, identified by visual inspection, include genes overexpressed in one or more tissues of origin. From top to bottom, clusters C.8, C14, C.18 and C.29 display probesets with high activity in adipose tissue, liver, brain and skeletal muscle and heart, respectively.
Table 1.
Group sizes in paired tissues.
Fig 3.
Extreme expression level profiles.
Expression levels of the most variable probeset (202286_s_at), on top, and a low-variability probeset corresponding to housekeeping gene GAPDH (217398_x_at), at the bottom, across all tissues for which there are at least 20 replicates of untreated, normal and cancerous samples. Samples from the same tissue of origin are displayed together, grouped by disease status. The green dashed line represents the overall mean; blue and red solid lines show the mean of cancerous and normal groups, respectively, whereas cyan and pink solid lines describe their respective dispersion, given by the within group standard deviation.
Fig 4.
Expression level changes across tissue types and disease status.
Distribution of the number of groups for which there is a clear change in the expression level of the probeset. The values are quantified by having either the group mean minus the group standard deviation above the overall mean or the group mean plus the group standard deviation below the overall mean.
Table 2.
Genes mapping to the top 100 probesets.