[TOC]

PPanG: a precision pangenome browser enabling nucleotide-level analysis of genomic variations in individual genomes and their graph-based pangenome

PPanG is a precision genome browser with a new perspective of pangenome. Existing pangenome browsers focus on the variations of nucleotide sequences, including base mutations and structural variations. However, pangenomic studies of the variations at gene level (presence/absence variations (PAV) and gene structural variations (gSV)) just stay on statistics, and the details of variation still remain unknown. PPanG provides nucleotide-accurate visualization of both genome sequences and genome annotations, making it clear to analyze the genomic variations from the nucleotide level to gene level.

PPanG is composed of two subviews: graph view for the whole pangenome in SequenceTubeMap and linear view for each individual genome in JBrowse2. PPanG is easy to get started, but requires efforts to interpret the visualizations, as the visualizations are mixture of different views (graph & linear views) and different tracks (sequence & annotation tracks). So it is strongly recommended to understand the principles of PPanG first. The introduction page is https://cgm.sjtu.edu.cn/PPanG/.

User Guide

PPanG is divided into three areas: Navigation Area, Visualization Area and Functional Area, and sd-1 gene region is visualized by default: Next we will explain these three areas in details.

Navigation Area

Navigation Area allows users to provide the custom region for visualization besides the default sd-1 gene region. The navigation is very simple within three steps:

1. Select the target chromosome at Data. Pangenome graphs built by both MC and PGGB are available for users to choose.

2. Select the navigation type. There are three navigation types available:

   a) Navigation Type=built-in genes

   Just select from the built-in genes. e.g.The navigation of GS5 on chr05:

   

   b) Navigation Type=reference annotation gene ID

   Input the target MSU RGAP7 gene ID and select the gene ID from the list. e.g. The navigation of LOC_Os08g03060 on chr08:

   

   c) Navigation Type=custom region

   Input the target region by coordinates. Format: <sample name.chrxx>:<start coordinate>-<end coordinate>. In this step any individual can be selected as reference. The "Custom Path" box provides the list of available paths. Type and select the path name (related samples will be suggested if typing into the box), and "Region" is automatically filled with the selected path name. e.g. Data=chr11_mc, Navigation Type=custom region, Custom Path=CHAOMEO.chr11 and fill the region with CHAOMEO.chr11:6541923-6546025:

   

3. Click the "Go" button and wait a minute for loading. The visualization of Visualization Area will be updated.

There are several functional buttons beside "Go":

• Left: Shift the target region to (start - offset, end - offset). offset is half the length of target region.
• Zoom in: Double the scaling of visualization in graph view.
• Zoom out: Half the scaling of visualization in graph view.
• Right: Shift the target region to (start + offset, end + offset). offset is half the length of target region.
• Compress: Compress the visualization of graph view within the width of the window. This function may be useful because the whole pangenome graph can be too long to observe.
• Reset: Reset the translation/zoom scaling to default in graph view.
• Download Image: Download the visualization composed of graph view and linear views. Note that this function is not suitable for large graphs. For this use case, see FAQ for more details.

Visualization Area

The visualization of PPanG is composed of SequenceTubeMap graph view and JBrowse2 linear views. For interpretation of PPanG visualization, please refer to the PPanG homepage: https://cgm.sjtu.edu.cn/PPanG/. The custom options for visualization are available at Functional Area.

Graph View in SequenceTubeMap

In the graph view, the pangenome graph of nine representative genomes are visualized by default. The graph is ultra long because each nucleotide is visualized, and the graph can be translated by mouse-dragging, zoomed by mouse-scrolling and compressed by clicking the "Compress" button at Navigation Area.

Zoom Behavior

The zoom behavior of "Zoom" button at Navigation Area and mouse-scrolling is different, because sometimes the zoom is only required at horizontal direction (compress the graph):

• For "Zoom" button: It is equivalent to normal zoom scaling.
• For mouse-scrolling: If horizontal zoom scaling < 1.0, the zoom only works horizontally. That is to compress/uncompress the graph. If horizontal zoom scaling > 1.0, the zoom works both horizontally and vertically. That is equivalent to normal zoom scaling.

Coordinates

Above the graph is the coordinate axis of the reference (not necessarily the reference genome, but the reference selected in the target region). In SequenceTubeMap, the coordinate axis may be longer than the target region, and the actual start and end coordinates are marked in yellow circles. For example, in the figure below, the yellow circle represents the start coordinate 6,541,923:

Linear View in JBrowse2

By default, the linear view of reference is visible below the graph view. Linear views for other individuals are simply added by double-clicking the paths in graph view. The usage of linear view is the same as native JBrowse2 component. (The more detailed document of JBrowse2 is available at https://www.jbrowse.org/jb2/docs/.)

Combination of Graph View and Linear View

In PPanG, the graph view and linear views are combined in parallel. The connection among different views are shared coordinates. Different linear views are all aligned to the graph view according to start and end coordinates. Besides, the coordinate change in graph view will trigger synchronous changes in linear views (The opposite is unavailable, because the linear views are free to extend outside the target region while the graph view is not). Consequently, exons at the same position from different views represents the same exon. Note that the coordinate alignment cannot be exactly perfect, because the coordinates of graph view are not uniformly distributed with occasional variations.

Functional Area

Functional Area contains several additional features helpful to users:

BLAT server

As the list of reference gene IDs is available in PPanG, it is easy to navigate to the known gene region. But it does not work for novel sequences, including distributed genes absent in reference genome but present in other genomes. Therefore, a BLAT server is embedded in PPanG, which helps the locating of novel sequences. Input the target sequence and click the "Search" button, the BLAT server will automatically process the alignment result and provide the target region. By default, the novel sequence is searched within the nine reference genomes, and the search range can be expanded to all individuals by selecting the "Search all genomes". Then input the "tRegion" (target region) of BLAT result as Navigation Type=custom region in 2c to visualize the custom region in PPanG.

Annotation Data

The Annotation Data tab collects genome annotations within the target region. Click the "Download All" button to download these annotation data for each individual.

Legend

The legends of all aligned genome tracks in graph view are shown in this tab. By default, the nine reference genomes are selected to show, and users are able to select any individual to add it in the pangenome graph. To visualize all individuals, please click the "Select all" and wait a moment. The large pangenome graph may cost much time and memory.

Visualization Options

There are some options to adjust the visualization in PPanG.

For SequenceTubeMap view:

• Hide nucleotide bases: Hide the text of nucleotide bases for simplicity.
• Fully transparent nodes: Turn the nodes transparent. Otherwise, the nodes are covered with gray masks.
• Show exons on genome tracks: Visualize the exons on genome tracks. Otherwise, only genome tracks are visualized.
• Select genes of no interest to hide: Hide exons from the selected transcripts. It is useful when different transcripts overlap and cannot be distinguished.
• Colors: By default, the genome tracks are colored in pale colors and exons are colored in grayscale. Users are able to select the color options to their preferences.

For JBrowse2 views:

• Rearrange linear views: Reorder or remove the linear views for better visualization.

Other options for JBrowse2 are available by clicking the button in the top-left corner of each linear view.

Quick Start

Example for MSU RGAP7 genes (reference annotation)

The visualization of reference annotation genes (e.g. LOC_Os08g03060) is simple with the steps below:

• Select "chr08_mc" at Data;
• Select "reference annotation gene ID" at Navigation Type and type the gene ID LOC_Os08g03060 into MSU RGAP7 Gene ID;
• Click the "Go" button.

Example for distributed gene Xa7 (novel sequence)

This section we will describe the whole user path for visualization of the distributed gene Xa7.

Xa7 is known to be absent in the reference genome, so the Navigation Type=reference annotation gene ID is unable for navigation. Then, the BLAT server is necessary to locate the Xa7 region. The Xa7 sequence is input into the BLAT server to find the target "tRegion": As the BLAT result shows, there is only one match with the "tRegion" of NATELBORO.chr06:28873554-28874897. Then the region is provided in the Navigation Area step 2c:

• Select "chr06_mc" at Data;
• Select "custom region" at Navigation Type and copy the tRegion NATELBORO.chr06:28873554-28874897 into Region;
• Click the "Go" button.

The visualization is shown below: As this figure shows, only one track is visible for nine reference genomes. That is, Xa7 sequence is only present in one genome and absent in the other eight genomes. All aligned tracks are listed in the Legend tab: Click the "Select all" button to visualize all individuals and click the "Compress" button at Navigation Area for a global overview. Double click some of the paths in graph view to add them into linear views. The final visualization is as follows:

Run PPanG for your own data

• Install

  clone the repo and install dependencies:

  git clone git@github.com:SJTU-CGM/PPanG.git
  cd PPanG/
  npm install # or yarn install

  tabix (https://github.com/samtools/tabix) and vg (https://github.com/vgteam/vg) are also needed in your PATH.

• Configuration for SequenceTubeMap

  The configuration of SequenceTubeMap view is in src/config.json, dataPath should be set to your own data folder (in PPanG dataPath is riceData/) and DATA_SOURCES correspond to the xg files in your dataPath. The reference is set in reference. The name, alias and annotation of reference can be the same. bedFile is only available if vg chunks are pre-processed, otherwise it should be removed. Other detailed configuration is available in SequenceTubeMap.

• Configuration for JBrowse2

  All genomes and GFF3 annotations are needed in bgzip format with tabix index (.fasta.gz, .fasta.gz.gzi, .fasta.gz.fai, .gff.gz, *.gff.gz.tbi) in jbrowse/ folder.

• Notes for BLAT server

  Unfortunately, the total size of all 113 genomes in PPanG is over the max limit of the BLAT server. So indeed, these genomes are divided into 11 parts of database. The BLAT server searches part1 (for nine references) or all parts (for all genomes) for the query sequence and collects the results together. That explains why searching all genomes may exceed the time limit. We do not recommend deploying the BLAT server as PPanG does. For novel sequences, just use sequence alignment tool like BLAST at user's local machine and ignore the BLAT server.

• Other Configurations for PPanG

  builtin_genes.json and reference_genes.json are required in src/ folder. The format is:

  {
  "<data.xg>": {
    "<gene_id>": "<gene_region>",
    ...
  },
  ...
  }

  For example, the builtin_genes.json in PPanG is like:

  {
  "chr01_mc.xg": {
    "sd1 (LOC_Os01g66100)": "IRGSP-1.0.chr01:38382381-38385503",
    "Pish (LOC_Os01g57340)": "IRGSP-1.0.chr01:33141126-33145608",
    "Gn1a (LOC_Os01g10110)": "IRGSP-1.0.chr01:5270102-5275677"
  },
  "chr02_mc.xg": {
    "tgw2 (LOC_Os02g52550)": "IRGSP-1.0.chr02:32155212-32156481",
    "GW2 (LOC_Os02g14720)": "IRGSP-1.0.chr02:8114960-8121924",
    "OsGL1-4 (LOC_Os02g40784)": "IRGSP-1.0.chr02:24718046-24724119",
    "tms5 (LOC_Os02g12290)": "IRGSP-1.0.chr02:6397341-6399235",
    "EP3 (LOC_Os02g15950)": "IRGSP-1.0.chr02:9042075-9046140",
    "RSS1 (LOC_Os02g39390)": "IRGSP-1.0.chr02:23769389-23772841",
    "OsSKIPa (LOC_Os02g52250)": "IRGSP-1.0.chr02:31995217-31997696"
  },
  ...
  }

FAQ

Q: Why sometimes the linear view reports the bug like "failed to fetch data, reload"?

A: This bug occurs possibly due to the network failure. Click the "reload" button will solve the problem.

Q: How to download the large pangenome graph?

A: The library used by "Download Image" button is designed for daily use, when users hardly need to download a so large graph. It is not suitable for the large pangenome data. As an alternative, we recommend a better way to download the large graph with a native tool in the Edge/Chrome browser. Press F12 to open the developer tools. Select the Elements tab, double-click the <div id="root">, and double-click the <div> below. Then, right-click the <div id="Pangenome browser">...</div> and click Capture node screenshot. Then the browser will automatically download the large graph into xx.png.

Q: Is it unavailable to search all genomes in the BLAT server if it exceeds the time limit?

A: Actually, the BLAT server uses caches to store the alignment results. Although one request exceeds the time limit, the caches are still saved in the server. If the same request comes, the BLAT server will use cashes to avoid redundant calculations. So it is feasible to click the "Search" button again and again until success.

Q: Why one genome has more than one paths in some regions in the graph view?

A: From the perspective of the whole pangenome graph, one genome should correspond to only one path exactly. But in a directed cyclic graph, one path may be truncated into different pieces within a target region. We show a simple demo in the figure below: There are totally two paths in this graph. However, if the right part of this graph is extracted as marked in the figure, there will be many broken lines in the subgraph. Actually, how to extract a pretty subgraph from the original graph is also a challenge to pangenome graph algorithms.