URL: https://sysimm.ifrec.osaka-u.ac.jp/immuno-navigator/
Publication: “Immuno-Navigator, a batch-corrected coexpression database, reveals cell type-specific gene networks in the immune system” Paper in PNAS or PubMed
Updates and other information will also be announced on twitter @ImmunoNavigator.
Introduction
The Immuno-Navigator database is a large collection of cell type-specific gene expression and co-expression data for cells of the immune system. It is a useful dataset for analysis of gene expression and coexpression networks.
At present, Immuno-Navigator contains co-expression data based on 4,639 human samples, obtained from 19 cell types from 191 studies, and 3,434 mouse samples, obtained from 24 cell types from 261 studies in total. In contrast with existing databases, Immuno-Navigator provides coexpression data in a cell type-specific way. A second difference lies in the processing of the gene expression data for batch effects, prior to the calculation of correlation data. In our paper, we showed that the processing of batch effects strongly improved the estimated correlation of expression.
Example usages:
Single-gene exploratory analysis
Searching for your gene of interest (for example Rela or Foxp3) allows the user to see cell type specific gene expression patterns (tab “Probes”), and top correlated genes in each cell type (tab “Top correlated genes”). In addition to predicted regulatory motifs in the gene’s promoter (tab “Predicted motifs”), you can inspect enriched motifs in the promoters of correlated genes (tab “Motif enrichment”). The same can be done for gene ontology terms (tabs “Gene Ontology Annotations” and “Gene Ontology Enrichment”). In this way, using the “guilt-by-association” principle, hypotheses can be made about the regulation or function of genes.
Gene pair correlation analysis
The expression of any pair of genes can be directly visualized in any cell type through the Gene Pair Comparison function. For example, the expression of Rela (a subunit of NF-κB) and Il6 in mouse macrophage samples is highly correlated (Fig. A below). This, together with the predicted NF-κB binding site in the Il6 promoter (Fig. B) and the enrichment of this motif in genes with correlated expression with Il6 (Fig. C), suggest that Il6 is indeed under the control of NF-κB.
Visualization of small correlation networks
The correlation of expression of small sets of genes in a cell type of choice can easily be visualized using the Correlation Network Creator function.
More complex analyses Correlation GSEA can be used to see if a single query gene has a high correlation with a set of genes compared to the genome-wide set of genes. High correlation with a set of genes might reflect a regulatory interaction. This function can also be used to select correlated genes from a set of candidate target genes (e.g. ChIP-seq based), resulting in a higher-confidence set of target genes (i.e. not only bound by a regulator, but also correlated in expression with the regulator).
Using Correlation Network Hub Prediction, a user can find genes that are highly connected with a set of input genes in the correlation networks. Such highly connected genes might include potential regulators.
Data download:
The following data is available for bulk download (see Download):
- Gene expression data is available for all 19 human and 24 mouse cell types in tab delimited tables.
- Gene coexpression data is availablefor download as (large!) binary files. Example scripts showing how to read these data are also available.
- Sample-to-cell type and batch annotation data. We used studies as proxies for batches.
Example screenshots: (A) Rela and Il6 expression is correlated in mouse macrophage-derived samples. (B) The Il6 promoter contains a predicted Rela binding site. (C) The promoters of genes with high correlation of expression with Il6 in macrophages are enriched for Rela binding sites.
