Single Cell Portal now supports exploring multiome ATAC data!  Upload your 10x Genomics multiome data to SCP to explore how your sample's chromatin accessibility aligns with gene expression.  Simply add the ATAC fragment data from your analysis pipeline, then you can view ATAC data in genomic context, and interrogate it with SCP's filtering and other powerful exploration features.

This 30-second animation shows how to filter ATAC data in the genome browser.

Let's dive into what this animation shows, and how we can extend that exploration to integrate gene expression and biological pathways.
 

Example: exploring an unannotated dataset

The Single Cell Portal study below shows chromatin accessibility around the SPIB gene in SCP2699.  This example derives from the public 10x Genomics dataset "PBMC from a Healthy Donor - No Cell Sorting (3k)".

A BED track at bottom in the genome browser shows ATAC reads, each grey or red line representing an ATAC-seq fragment.  More reads in a genomic region means more accessible DNA.  These areas are primed for transcription.

Assess ATAC read quality at a glance

Click "-" at top right in the genome browser to zoom out a bit, then click on a red ATAC read.  These features are colored by ATAC read count.  Higher read counts (red) can occur due to technical factors with the data, like PCR duplicates, sequencing errors, or low complexity regions.  We can see at a glance that most ATAC reads are grey, having a read count of 1 and thus more reliable data.  You can click on the ATAC features to see read count and other metadata.

Filter ATAC reads by raw annotations

This public dataset has "raw" annotations -- placeholder labels like "0", "1", and "2" for gene expression (GEX) and ATAC clusters.  Click "Filter plotted cells" to explore these annotations -- see green arrow at right in Figure 1 above.  In the resulting filter panel, click "More…", then hover over "Atac cluster" and deselect the checkbox that appears next to it to hide all ATAC reads in the genome browser.  Then click "6" in the column of checkboxes, and note few ATAC reads appear.  That suggests this region has little accessible chromatin for cells in ATAC cluster 6.

Now click the "7" checkbox.  Lots of ATAC reads appear just upstream of the 5' end of SPIB, where we would expect the gene promoter!  This may imply that cells in ATAC cluster 7 are poised for expression of SPIB.

Compare chromatin accessibility and gene expression

Having filtered ATAC reads in the genome browser, we suspect that SPIB in ATAC cluster 7 has high chromatin accessibility and is thus poised for expression.  But is it indeed expressed in those cells?  And which GEX clusters of cells is it expressed in?  

To answer this question, keep your filter selection as-is, and click the "Distribution" tab at top of the page.  This shows raw labels for the gene expression cluster (gex_cluster) annotation.  Looking at the violin plot, we can immediately see the SPIB gene is expressed in GEX clusters 1 and 5 with higher levels of expression in GEX cluster 5.  

So we can conclude that SPIB expression in GEX cluster 5 corresponds to SPIB chromatin accessibility in ATAC cluster 7.

Comparing the annotations “gex_cluster” and “atac_cluster” annotations in the scatter plot confirms that GEX cluster 5 and ATAC cluster 7 represent the same cell population.

Explore related genes and pathways

We've seen evidence that the SPIB gene is expressed in GEX cluster 5 and ATAC cluster 7.  We might like to know a little more about the gene, which could help generate useful hypotheses.  What tissues is SPIB typically expressed in?  What genes is SPIB known to interact with, and in which biological pathways?  How does the expression of genes look in those pathways -- in this dataset?

To explore these questions, let's look over the genes plotted along chromosomes above the violin plot.

Hover over the red "SPIB" gene label.  The popover reports SPIB is a transcription factor.  It also shows that SPIB is disproportionately expressed in lymphocytes (B cells, T cells, and natural killer cells) per reference data in GTEx.

Hover over the purple "TLR9", and note the tooltip reports this gene is also highly expressed in lymphocytes.  Now click the pathway link "Transcriptional regulation of memory B cell differentiation" to see biochemical context on how SPIB and TLR9 interact.

The resulting pathway diagram shows a gene regulatory network in cellular context.  Arrows represent the direction of molecular action.  Taken together, they show causal cascades from upstream genes to downstream, higher-level effects.

As you can see upon hovering over the blue "i" icon, red means higher scaled mean expression, blue means lower, and fainter whiter color means lower percent of cells expressing.  (Text color is simply set to the most readable visual contrast.)  The halo outline color corresponds to whether the gene is the searched gene (red outline) or the interacting gene you selected (purple outline).  Changing the "Expression in" dropdown menu from "1" to "5", changes the gene expression cluster visualized in the pathway diagram. Recall in GEX cluster 5 we saw SPIB was overrepresented in the expression violin plot and the genome browser's ATAC reads track.

The resulting pathway overlay shows notable expression in genes related to B cell receptor signaling.  For example, CD79A (bold red) has a high percent of cells expressing and high scaled mean expression, and  SPIB (faint red) has medium percent and high mean.  The diagram also shows overall low expression in pathway nodes for development in the germinal center -- a structure in lymph nodes and spleen where immature B cells enter, then proliferate and differentiate.  For example, BCL6 (faint blue) has low percent of cells expressing and low scaled mean expression.  With these dataset-specific expression summaries shown in the context of biochemical pathways, we can refine hypotheses about cell type and cell state.

Upload your multiome data

Got multiome ATAC data you want to analyze and share, like the example discussed above?  Simply create a study, upload your expression data in SCP "Classic" format, and add your ATAC fragment data as an indexed BED file in the upload wizard.

The Single Cell Portal team wants feedback on how we could improve support for multiome ATAC!  Email us at support@broadinstitute.zendesk.com with suggestions.