HOMER Motif Enrichment Analysis in SciDAP

Transcription Factor motifs

Motif analysis is used to identify short sequences (motifs) that are over-represented in target sequences compared to background. These can be binding sites for sequence-specific DNA or RNA-binding proteins such as transcription factors (TFs). Researchers may use it for several applications such as: 1) identifying a motif for a new TF based on ChIP-Seq data, 2) confirming that ChIP-Seq worked by showing that the motif for the TF is enriched in ChIP-Seq peaks, and 3) identifying candidate TFs that bind a set of promoters or enhancer elements identified by differential expression or accessibility (ATAC) analysis. HOMER, a widely-used tool in this field, features a discovery algorithm tailored for genomics applications [1], [2]. Employing a differential motif discovery approach, HOMER pinpoints TF binding sites enriched in one set of sequences compared to another.

Screenshot of SciDAP’s HOMER motif analysis workflow output.

To appreciate the significance of HOMER’s workflows, it’s essential to understand a brief background of motif analysis. Motifs are short, recurring sequences that play a crucial role in gene regulation by serving as binding sites for TFs. Identifying these motifs can unravel the intricate mechanisms governing cellular functions. HOMER’s algorithm, designed with ChIP-Seq and promoter analysis in mind, has become instrumental in deciphering these regulatory elements.

Motif analysis can be used to identify known or de novo motifs. When searching for known motifs, the HOMER pipeline will calculate enrichment for a limited number of TF binding sites taken from a TFBS database. For this reason, multiple members of a TF family will be reported if present in the database. For example, if AP-1 motif is enriched, JUN, FOS, BATF and ATF factors will be listed separately. In most cases, motif analysis cannot distinguish between members of the TF family. For de novo analysis, HOMER will look for any sequence over-represented among targets which will take much longer. Then, the enriched motifs will be compared to the TFBS database and the top matching motif will be reported. In this case, if the AP-1 motif is enriched, only one of the AP-1 proteins will be in the report.

Motif Finding with HOMER and Random Background Regions

When calculating the enrichment, it matters what sequences are used as a background. HOMER can use either random or custom background depending on the scientific question. If the random background pipeline is selected, Homer will use neighboring sequences to create a background set of sequences. Beginning with a target file containing target regions denoted by “chr,” “start,” and “end,” the data undergoes refinement, sorting, and duplicate removal. Subsequently, each region is extended by 20 kilobases in both directions to capture potential regulatory elements. Merging overlapping extended regions creates a comprehensive genomic set. After subtracting the target regions, the remaining sequences are used to generate the background.

Motif Finding with HOMER and Custom Background Regions

Instead of relying on a randomly generated background, users may define a custom set of background regions based on their research context. This provides more control over the background selection process, allowing for targeted analyses aligned with the study’s biological context. For example, if a scientist wants to identify the TF that opens chromatin following a treatment with a drug, they may conduct ATAC analysis to identify the regions that opened. These regions will be the target set. In this case, all open regions in the control sample will be a good background which will avoid enrichment for cell-type specific TF motifs.

How to create the lists of target and background regions?

SciDAP allows users to create lists of regions based on their data. Users can make lists by filtering the peaks called in ChIP/ATAC or CUT&RUN experiments. Alternatively lists of peaks can be created from differential analysis such as DiffBind. Users can further refine their lists by filtering on any column of output data, or by comparing/contrasting lists using the set operations workflow that enables producing a union, intersection, difference, or complement of any number of lists [3].

Example 1. Did my ChIP-Seq or CUT&RUN work?

To confirm that a ChIP-seq experiment worked one may want to conduct de novo motif analysis to identify the motif enriched in the ChIP-seq peaks. For this, one can select strong, but not the strongest peaks called in the ChIP-Seq experiment by p/q-value or enrichment value. Avoiding the strongest peak helps to avoid peak artifacts. We prefer to select 1000 peaks ranked 500 to 1500 by q-value for this analysis.

1. You’ll need to first go to your ChIP-Seq sample in SciDAP and click on the “Peak Calling” result tab. Sort this table by the “log10q” column.
2. Note the log10q value in row 500 and row 1500. These log10q values will be your maximum and minimum values, respectively, when filtering this sample using the filter peaks workflow (workflow [e], below).
3. Under the “GeneList” tab on the project page, select “ADD GENELIST” then select workflow [e], assuming it has already been added to your project. Fill out a name for the new GeneList and select the sample you want to filter.
4. Under the first “Filtering parameters” set of drop downs in the filter peaks workflow, select “log10q”, “Less Than Or Equal”, and fill in your maximum log10q value. Then click the “+” sign to add another set of parameter drop downs. For this second set, select “log10q”, “Greater Than Or Equal”, and fill in your minimum log10q value here. Be sure to have “And” selected above the first parameter set before clicking “Save sample”.
5. Once the filtered sample is completed, you can use this genelist as input into the HOMER Random background workflow (workflow [c], below).

Screenshot of SciDAP’s Filter peaks workflow input form. The (+) button has been selected, the “AND” operatoris set, and the user is in the process of selecting a second “log10q” filtering parameter from the dropdown list.

Screenshot of SciDAP’s Filter peaks workflow input form. The (+) button has been selected, the “AND” operatoris set, and the user is in the process of selecting a second “log10q” filtering parameter from the dropdown list.

Example 2. What TF factors changes epigenome?

To identify a TF that opens chromatin following a drug treatment, scientists can conduct ATAC in control and treated cells. Diffbind can be used to identify the sites where chromatin opens or closes upon treatment. The output of diffbind can be filtered to identify the sites that change significantly (e.g., FDR<0.1 and log2 Fold change >1) to create the target peak set.

1. After running the Diffbind workflow on your two groups of cells, you will be able to use it as input for the Diffbind peak filtering workflow (workflow [a], below).
2. Under the “GeneList” tab on the project page, select “ADD GENELIST” then select workflow [a], assuming it has already been added to your project. Fill out a name for the new GeneList and select the sample you want to filter.
3. Under the first “Filtering parameters” set of drop downs in the filter peaks workflow, select “FDR”, “Less Than Or Equal”, and type “0.10” into the bottom field. Then click the “+” sign to add another set of parameter drop downs. For this second set, select “Fold”, “Greater Than Or Equal”, and type “1” into the bottom field. Be sure to have “And” selected above the first parameter set before clicking “Save sample”.
4. Once the filtered sample is completed, the resulting peak list is of all open sites in your control cells, and can be used as a custom background for the HOMER Custom background workflow (workflow [d], below).

Screenshot of SciDAP’s Filter Diffbind workflow input form. The (+) button has been selected, the “AND” operator is set, and the user is in the process of selecting a second filtering parameter from the dropdown list.

Example 3. My target list is not from SciDAP

Within the two HOMER workflows [c][d], users can also provide region lists created outside SciDAP by selecting “Skip Upstream”. When this is enabled, a File Manager button will appear that can be used to upload the file containing peak coordinates. The file should be BED formatted with tab separated variables, and contain three columns: chr, start, and end. These fields must be string, integer, and integer data types, respectively. Finally, this file should not have a header as the first row.

Motif Analysis

To start your motif analysis in SciDAP, please follow this quick-guide:

1. Assuming you’ve already registered an account and have a project with ChIP-Seq, ATAC-Seq, or Cut & Run samples, you’ll need to add the following Global workflows to your project:

a. Filter differentially bound genes from DiffBind for heatmap analysis

b. Set Operations for filtered genelists

c. Motif Finding with HOMER with random background regions

d. Motif Finding with HOMER with custom background regions

e. Filter ChIP/ATAC peaks for Tag Density Profile or Motif Enrichment analysis

2. You’ll first need to run a genelist workflow as described above in Examples 1 and 2, or prepare your own bed file as input. Depending on your research objectives, you may also want to run a set operation between lists using workflow [b].

3. Now that you have your list of peaks samples, use these as input to either workflow [c] and/or [d] to analyze the DNA sequences underlying the peak data for motifs relative to a random or custom background.

That’s it! Now biologists can easily perform motif analysis using HOMER without learning to code or running command line tools. With the SciDAP genelist and HOMER workflows, you can move beyond the command line to easily identify differences in gene expression and gene sets between experimental phenotypes. Try it for yourself with a FREE trial.

References:

1. Heinz S, Benner C, Spann N, Bertolino E et al. Simple Combinations of Lineage-Determining Transcription Factors Prime cis-Regulatory Elements Required for Macrophage and B Cell Identities. Mol Cell 2010 May 28;38(4):576–589. PMID: 20513432
2. http://homer.ucsd.edu/homer/motif/
3. Shane Neph, M. Scott Kuehn, Alex P. Reynolds, et al. BEDOPS: high-performance genomic feature operations. Bioinformatics (2012) 28 (14): 1919–1920. doi: 10.1093/bioinformatics/bts277