In RNA-Seq data analysis, once a set of differentially expressed genes has been obtained, the next important step is to interpret those changes biologically.
In the RNA-Seq Data Analysis Tutorial for Subio Platform,
we introduced enrichment analysis using predicted transcription factor target gene sets provided by GSEA in
“Chapter 7: Gene Annotation and Enrichment Analysis: From Statistical Results to Biological Interpretation.”
We also showed how to search for genes that have a specific motif sequence within a user-defined distance from the transcription start site (TSS) in
“Chapter 8: Genes Regulated in a Genome Position-Specific Manner and Motif Sequences.”
However, both of these analyses are based on sequence information or predicted gene lists. The presence of a motif sequence does not necessarily mean that the transcription factor actually binds there in the relevant cell type or experimental condition. Predicted target gene lists also do not directly show the actual binding status or transcriptional regulation.
In this article, we use ChIP-Atlas, a database that integrates public ChIP-seq data, to show how RNA-Seq expression patterns can be interpreted in relation to transcription factor binding information.
What is ChIP-Atlas?
ChIP-Atlas is a data-mining suite that integrates experimental data such as ChIP-seq, ATAC-seq, and Bisulfite-seq registered in public databases, allowing users to explore epigenomic information. On the ChIP-Atlas top page, it is described as integrating more than 430,000 ChIP-seq, ATAC-seq, and Bisulfite-seq experiments.
ChIP-Atlas provides several functions, including Peak Browser, Target Genes, Enrichment Analysis, Colocalization, and Dataset Search. In this video, we mainly use the following two functions.
| Function | Main purpose | How it is used in this video |
|---|---|---|
| Peak Browser | To examine peaks of specific TFs, histone modifications, and other features on the genome | To obtain ESRRA ChIP-seq peaks as a BED file and import them into Subio Platform |
| Enrichment Analysis | To examine which TFs and other features have many peaks near an input gene set or genomic region set | To search for TF candidates whose binding peaks are frequently found near upregulated genes |
How should ChIP-Atlas results be interpreted?
ChIP-Atlas allows us to go beyond motif-based prediction and use transcription factor binding peaks observed in actual ChIP-seq experiments as clues for considering upstream regulators of gene expression changes.
However, ChIP-Atlas results need to be interpreted carefully. Transcription factor binding can vary greatly depending on cell type, stimulus, differentiation state, time point, and chromatin state. Therefore, the presence of a peak in a region means that binding was observed in a particular past experimental condition. It does not mean that the transcription factor always binds to that region.
In addition, because ChIP-Atlas integrates public data, the antibodies, sequencing depth, peak detection sensitivity, and data quality of the included experiments are not uniform. If a peak is found in ChIP-Atlas, it can be a useful clue that the transcription factor may be involved. On the other hand, even if no peak is found, it may simply mean that data for the relevant cell type or condition have not been registered, or that the binding was not detected because of experimental sensitivity or data quality.
In other words, ChIP-Atlas is very useful for finding evidence that a binding event has been observed, but it is not a tool for proving that binding does not exist.
Obtaining ESRRA binding peaks with Peak Browser
In the first half of the video, we use the ChIP-Atlas Peak Browser to download ESRRA binding peaks detected in breast-derived cells as a BED file.
The downloaded BED file can be imported into Subio Platform as a Region List. By importing it as a Region List, the genomic positions of transcription factor binding peaks can be examined together with RNA-Seq data and gene annotations.
In this example, we narrowed the ESRRA peaks down to those detected in experimental data using MDA-MB-231 cells. We then visualized the peak positions on the hg38 genome and used Genome Location Filter to extract genes that have ESRRA peaks near their TSS.
As an example, we extracted genes that have peaks in a region centered on the 500 bp upstream region of the TSS, with an additional 500 bp extension on both sides. In this way, Subio Platform allows users to search for genes that have peaks in specific genomic regions by specifying positions relative to the TSS.
Examining expression patterns of genes with ESRRA peaks
The extracted gene list can then be used directly to examine expression patterns in the RNA-Seq data. When expression data are visualized in the Genome Browser, genes located relatively close to each other may appear to show similar expression patterns.
This can provide clues for considering not only ESRRA-mediated transcriptional regulation, but also epigenetic regulation related to chromatin state or genomic position. Of course, this does not prove causality at this stage. However, by examining genomic position, transcription factor binding peaks, and expression patterns on the same screen, it becomes easier to decide which gene groups should be examined in more detail.
Searching for upstream candidate TFs with Enrichment Analysis
In the second half of the video, we take the opposite approach. That is, starting from a gene group showing a specific expression pattern, we search for candidate transcription factors whose binding peaks are frequently found near those genes.
Here, we extract genes that are commonly upregulated by three ESRRA siRNA treatments, and input their Gene Symbols into ChIP-Atlas Enrichment Analysis. Enrichment Analysis is a function that examines which transcription factors or epigenomic features have many peaks overlapping with, or located near, the input gene set or genomic region set.
This is not an analysis for checking whether a specific TF peak exists for one particular gene. Rather, it is an analysis for a gene group showing a particular expression pattern, used to examine which TF binding peaks are frequently found nearby and to search for transcription factor candidates that may be involved upstream.
Checking Enrichment Analysis results in Subio Platform
ChIP-Atlas Enrichment Analysis results can be downloaded as a TSV file. In the video, we open the result in Excel and filter it for experimental data from MDA-MB-231 cells. We then extract the Gene Symbol column, save it as a tab-delimited file, and import it into Subio Platform as a gene list.
When these upstream candidate TF genes are clustered in Subio Platform, ESRRA is found. Because ESRRA expression is reduced by ESRRA siRNA treatment, while genes with ESRRA binding peaks nearby are upregulated, this provides a clue that ESRRA may normally suppress the expression of these genes.
Why combine ChIP-Atlas with RNA-Seq?
RNA-Seq analysis alone can tell us which genes are upregulated or downregulated, and which pathways or GO terms may be involved. However, to consider which transcription factors may be upstream of those expression changes, another type of information is needed.
With ChIP-Atlas, we can use public ChIP-seq data to examine which TF binding peaks are frequently found near differentially expressed genes. This makes it possible to connect expression patterns found by RNA-Seq with information about transcription factor binding and chromatin state.
ChIP-Atlas is a powerful tool for hypothesis generation, helping narrow down candidates that should be examined in the next step.
Video