For this week's exercise, we will be constructing a pangenome graph using the Minigraph-Cactus pipeline. We will focus on a small region of chromosome 8, with sequences pulled from the T2T-CHM13 linear reference and from the phased assembly of HG00621, one of the individuals used in the Human Pangenome Reference Consortium (HPRC) pangenome reference.
First, make a working directory. Download the following 2 files and place them in this directory:
Next, make a directory called data within the working directory, and place the following 3 files in that directory:
Now, from the top level of the working directory (not inside the data directory), we can launch our construction docker container:
docker run -it --rm --cpus 1 --memory 8589934592 --memory-swap 8589934592 -v `pwd`:/data mgibio/cactus:2.5.0-focal-legacy /bin/bash
Inside the container, switch to our working directory: cd /data. Activate the virtual environment configured for minigraph cactus: . /cactus-bin-v2.5.0/cactus_env/bin/activate. Finally, start the alignment and graph construction by running the provided script: bash mcgb.sh
This may take 10-15 minutes to run. After the script completes, you can exit the container. All outputs will be found in the out directory.
The provided script launches the minigraph-cactus pipeline. There is a wrapper command that runs the entire pipeline (cactus-pangenome), but due to some occasional bugs, we find it more reliable to manually run each command in the pipeline. Let's go through a high-level overview of each step:
cactus-minigraph: this uses minigraph to progressively build an initial graph, starting from the reference assembly and aligning large syntenic chunks from each additional assembly in the order provided. The resulting graph will only contain large SVs.cactus-graphmap: this uses minigraph to map each assembly back to the graph constructed in the previous stepcactus-graphmap-split: this splits the assemblies and the mappings from the previous step by chromosome. Optional, but reduces memory in the next steps, which is especially important for our purposes, since we are running this exercise locally with limited RAM.cactus-align: this combines the mappings from the previous step into a multiple genome alignment, then converts that into a cactus graph. See the Minigraph-Cactus paper for more information about these structures.cactus-graphmap-join: this step runs several post-processing steps, such as normalizing, clipping, and filtering, to produce the final output graph. Also produces indices. The exact processing steps performed and indices generated may be adjusted based on a variety of flags as appropriate for the desired downstream analyses.
Note that some of these commands are themselves wrappers around multiple steps. For more information, see the Minigraph-Cactus documentation and the Minigraph-Cactus paper.
Look at the outputs in the out directory. Which files are indices, and which are graph files? What are some of the differences between the various graph file types?
- Hint: Try looking for the file extensions in:
- The Minigraph-Cactus documentation
- VG file format list- a fairly comprehensive overview of graph-specific file types
- The "File Formats" and "Index Types" pages linked at the top of this page may also be useful- they describe some of the common formats in more detail
- Note that
.gzand.tgzextensions mean that a file has been compressed; you can ignore them when trying to determine the type of a file from its extensions
Download Bandage.
Unzip the file we will be loading: gunzip -c bschr8.gfa.gz > bschr8.gfa
Open Bandage, navigate to File > Load graph, and select bschr8.gfa
NOTE: Mac users- Your Mac may refuse to open Bandage with a warning that it can't check for malicious software. Bandage is safe to use- you can override this by clicking the Apple logo at the top left corner of your screen, then System Settings, then Privacy & Security in the sidebar. Scroll down to the Security section, and click Open Anyway next to Bandage. You should now be able to open it normally.
Under Graph drawing on the left tool bar, select Entire graph from the drop own, then click Draw graph. This may take a few minutes to load. Note that the window may appear to show nothing after loading is done, but that's just because it's not centered on any portions of the graph.
Once the graph has loaded, under Find nodes on the right tool bar, enter 8737 next to Node(s), then click Find node(s).
Use the + and - keys to zoom in/out, and the scroll bars in the window or the scroll on your mouse to move around. You can also click and drag nodes (the colored line segments, representing sequences) to change their position.
What seems to be happenening at this location? What would this look like in a linear reference, and what effect could this have on alignments at this locus?
Please send in a screenshot from Bandage of the region surrounding node 8737, as well a brief answer to the questions in the Graph Construction and Visualization sections.