A global map of human gene expression

Abstract

Although there is only one human genome sequence, different genes are expressed in many different cell types and tissues, as well as in different developmental stages or diseases. The structure of this 'expression space' is still largely unknown, as most transcriptomics experiments focus on sampling small regions. We have constructed a global gene expression map by integrating microarray data from 5,372 human samples representing 369 different cell and tissue types, disease states and cell lines. These have been compiled in an online resource (http://www.ebi.ac.uk/gxa/array/U133A) that allows the user to search for a gene of interest and find the conditions in which it is over-or underexpressed, or, conversely, to find which genes are over-or underexpressed in a particular condition. An analysis of the structure of the expression space reveals that it can be described by a small number of distinct expression profile classes and that the first three principal components of this space have biological interpretations. The hematopoietic system, solid tissues and incompletely differentiated cell types are arranged on the first principal axis; cell lines, neoplastic samples and nonneoplastic primary tissue-derived samples are on the second principal axis; and nervous system is separated from the rest of the samples on the third axis. We also show below that most cell lines cluster together rather than with their tissues of origin.The widely used GNF Gene Expression Atlas 1,2 includes a variety of normal tissue and cell types as well as certain disease states. Many more different biological states, such as rare diseases or particular cell subtypes, exist. It is impractical for a single dedicated experiment to generate a comprehensive expression data set covering all biological conditions, partly owing to cost, but also because some conditions are studied only in specialized laboratories. Even so, we can use computational approaches to integrate the wealth of experiments that already have been performed.Integration of independent microarray studies is challenging, as microarrays do not measure gene expression in any absolute units. Several studies have integrated single-platform 3 and cross platform 4-6 data from single-channel oligonucleotide arrays yielding consistent results. It has been generally accepted, however, that only data from the same platform can be reliably integrated on a quantitative level 7 . Integration is also challenging because of the unavoidable complexity of sample descriptions. The Unified Medical Language System has been used to re-annotate free text-based sample descriptions 8 ; however, extracting information from brazma@ebi.ac.uk .

HEMO, an ancestral endogenous retroviral envelope protein shed in the blood of pregnant women and expressed in pluripotent stem cells and tumors

“…To get insight into the possible expression of the HEMO gene in human tumors, we performed an in silico analysis of microarray data using the dataset E-MTAB-62 elaborated in ref. 34, which includes 1,033 samples from normal tissues and 2,315 samples from neoplasm tissues obtained from various ArrayExpress (AE) and Gene Expression Omnibus (GEO) studies (SI Methods). In normal tissues, as expected from the qRT-PCR analysis in Fig.…”

Machine Learning Approaches Distinguish Multiple Stress Conditions using Stress-Responsive Genes and Identify Candidate Genes for Broad Resistance in Rice

“…Metaanalysis of microarray studies is widely used, especially in clinical research, to improve statistical robustness and detect weak signals (Liu et al, 2013;Rung and Brazma, 2013). For instance, thousands of samples belonging to hundreds of cancer types were combined, which provided new insights into the general and specific transcriptional patterns of tumors (Lukk et al, 2010). Microarray studies are burdened with a high dimensionality of feature space, also called the "curse of dimensionality" (i.e.…”

Redefining the relevance of established cancer cell lines to the study of mechanisms of clinical anti-cancer drug resistance

“…Based on both in vitro models and clinical studies, the literature is replete with hundreds of prognostic and predictive markers, yet clinical progress in improving cancer treatment has been incremental at best (2). Besides issues associated with the limitations of technology and the selection of patients (3), the clinical relevance and the usefulness of in vitro models for assessing new therapies is controversial (4)(5)(6)(7)(8).…”