Human pathology


The Pathology Atlas is built on gene expression data that includes quantitative transcriptomics data (RNA-Seq) and spatial proteomics data (immunohistochemistry on tissue microarrays). The Pathology Atlas focus is on cancer tissues, corresponding to a major part of clinical pathology and the fundament for cancer diagnostics.

All transcriptomics data has been retrieved from the Cancer Genome Atlas and all proteomics data has been generated in-house using the same antibodies as in protein expression profiling in normal human tissues.


THE CANCER PROTEOMESi

In the Pathology Atlas we have used a systems level approach to analyze the proteome of 17 major cancer types with respect to clinical outcome based on genome-wide transcriptomics analysis of almost 8000 individual patients with clinical meta data. Gene expression data on both the mRNA and protein level corresponding to the 17 major forms of cancer shown below is presented in a gene-centric manner for all protein encoding genes. In addition, immunohistochemistry-based protein expression is also shown for three forms of cancer lacking transcriptomics analyses: lymphoma, carcinoids and non-melanoma skin cancer.

SURVIVAL ANALYSIS AND PROGNOSTIC GENESi

The gene expression level (mRNA) for each gene has been analyzed in all patients and further correlated to the overall survival time for each patient. A large number of genes show differential expression within a given type of cancer and this has allowed for the unbiased identification of prognostic genes, both favorable genes, where high mRNA expression correlates with long survival time and unfavorable genes, where high mRNA expression correlates with shorter survival times. For 6833 genes, high relative expression is correlated with poor prognosis in at least one of the analyzed cancer types while for 6113 genes, high relative expression is correlated with good prognosis in at least one of the analyzed cancer types.

More than 100 million Kaplan-Meier plots have been analyzed to find the correlation between gene expression levels and patient survival times. The median expression level has been used as cut-off for high and low expression together with the cut-off that provides the highest level of significance in the Kaplan-Meier analyses. All Kaplan-Meier plots corresponding to median expression cutoff and best separation cutoff are presented for each gene in each cancer type, corresponding to a total of 900 000 Kaplan-Meier plots. The number of significant prognostic genes in each cancer are listed below.

Note that several well-known cancers gene markers do not show significance here when patient survival and gene expression levels at time for diagnosis is analysed. An example of this is over expression of Her2 in breast cancers, no longer leading to shorter patient survival due to the introduction of Her2-specific biological drugs. Other well-known cancer markers are subject to activating mutations that result in altered proteins and these genes are not identified here since the mutations do not normally lead to increased RNA expression levels, e.g. K-Ras and p53.