SiSaNA can currently be found on bioRxiv
SiSaNA is a command line tool that utiliizes the PANDA and LIONESS algorithms from the netZooPy module to reconstruct single sample regulatory networks. Using SiSaNA, users can easily calculate in- and out-degree for each of the reconstructed networks. Additionally, SiSaNA can compare the expression/degree between groups of interest, including performing statistical tests, visualizing the results (volcano plots, boxplots, violin plots, and heatmaps), and compare the survival between groups. All this is accomplished via the command line, with little to no prior programming experience required.
PANDA is a tool developed to reconstruct gene regulatory networks from bulk input data (such as RNA-Seq data). It uses a message passing approach, along with prior protein-protein interaction data and prior regulatory knowledge information to refine a single network that reflects the regulatory landscape of the input samples. LIONESS utilizes the PANDA algorithm and iteratively removes one sample (with replacement), then uses PANDA to reconstruct a network with all samples minus one. It then uses differences in the original PANDA network and the newly created network with n-1 samples to reconstruct single-sample regulatory networks.
These networks consist of two types of nodes: transcription factors (TFs) and genes, with an edge that connects each TF to each gene. The weight (or value) of the edge denotes the likelihood of a TF to regulate that gene. A larger edge weight means a higher likelihood of regulation.
SiSaNA will also calculate in-degree and out-degree, where in-degree is the sum of edge weights coming in to a gene while out-degree is the sum of edge weights coming out from a TF. As you may have guessed, a larger in-degree means that, on average, a gene is being more highly regulated in the modeled disease. Meanwhile, a larger out-degree means that, on average, that TF is regulating more genes in the modeled disease.
- python v3.9.19 (see installation steps for creating a conda environment with this specific Python version). SiSaNA should work with versions of Python 3.9.0 or greater, but as it has been written and tested on this version, we will use 3.9.19.
- Create a conda virtual environment with python version 3.9.19. Note: You need to substitute the path you want on your own system for the --prefix argument
conda create --prefix </path/to/env-name> python=3.9.19
- Enter the conda environment
conda activate </path/to/env-name>
- Install SiSaNA via the pip package installer
pip3 install sisana
- Create a directory for the analysis and move into the analysis directory
mkdir sisana
cd sisana
Example input files can be obtained using the command
sisana -e
These files will be downloaded from Zenodo and stored in a directory called "example_inputs". One of these example files is the params.yml file, which can be used as a template and edited for your own data (see next section). Each user-defined parameter in the params.yml file is documented with a comment to explain the function of the parameter. The comments do not need to be removed prior to running SiSaNA. The files in this example_inputs directory can be used in the commands listed down below.
To view help documentation on which subcommands are available, the following can be used:
sisana -h
For further information on these subcommands, simply put the name of the subcommand before the -h
sisana <subcommand> -h
The most important thing to get right in order to correctly run SiSaNA is the structure of your params.yml file. SiSaNA comes with a params.yml file that is annotated to explain the function of each argument. The params.yml file is separated into 'chunks' that reflect the same subcommands available in SiSaNA on the command line. For each step of SiSaNA, you will need to use the correct subcommand, as well as have the parameters set up in the params.yml file.
In the below example, the user is running the "preprocess" step of SiSaNA. They have specified the paths to the input files as well as the value for the number of samples a gene must be expressed in (in their case, 5), along with the path to the output directory in which to store their results.
Important
We are aware of an issue on SiSaNA PyPI v1.5.3 where there is an issue that causes the error message "KeyError: 'visualize'" to occur if the user does not have a "visualize" section in their params.yml file. To address this issue, we are currently implementing a solution in the visualize_bug_fix branch. Alternatively, if you experience this error you can "fix" it by copying and pasting the "visualization" section of the example params file into your own params file (even though the parameter names may not match your own file paths). This error occurs when SiSaNA attempts to access parameters in the "visualize" section of the params file but the visualize section is not present. It should only occur in instances where individuals have deleted sections of the params.yml file that they were not intending to run for their analysis.
The "preprocess" subcommand is the first stage of SiSaNA, where it preprocess the input data to remove genes not expressed in enough samples (as defined by the user). Information regarding the removal of these genes is given at the end of the preprocessing step.
sisana preprocess ./example_inputs/params.yml
Three files, one for each of the three filtered input files.
This second SiSaNA stage, "generate", uses the PANDA and LIONESS algorithms of netZooPy to reconstruct gene regulatory networks. Documentation for netZooPy can be found at https://github.com/netZoo/netZooPy/tree/master. It then performs basic analyses of these networks by calculating in-degree of genes (also called gene targeting scores) and out-degree of transcription factors (TFs).
sisana generate ./example_inputs/params.yml
- lioness.npy, which contains all calculated edges for each sample
- lioness.pickle, which is the same thing, just serialized to make reading into python quicker
- A file containing the calculated indegree and another file with the outdegree of each gene and transcription factor, respectively.
The next stage in SiSaNA, "compare", is used to find out how the in-degree and out-degree differ between each group. SiSaNA offers multiple ways to do this comparison, including t-tests (and Mann-Whitney tests) and paired t-tests (and Wilcoxon paired t-tests).
To compare the values between two groups in order to identify differentially expressed genes or differential degrees, you can use the following command:
sisana compare ./example_inputs/params.yml
For performing survival analyses, you can use a command like this:
sisana survival ./example_inputs/params.yml
"sisana gsea" is used to perform gene set enrichment analysis (GSEA) to identify pathways that are differentially regulated based on the gene targeting scores. It uses the ranks of genes found in the previous step (sisana compare) as input.
sisana gsea ./example_inputs/params.yml
The "visualize" command allows you to visualize the results of your analysis on publication-ready figures. There are multiple types of visualization you can perform, including generating volcano plots...
sisana visualize volcano ./example_inputs/params.yml
...making boxplots or violin plots of expression/degrees...
sisana visualize quantity ./example_inputs/params.yml
...and creating heatmaps
sisana visualize heatmap ./example_inputs/params.yml
The final stage of SiSaNA, "summarize", takes all the created images and outputs them in a single html file for convenience. This can then be opened in a web browser. Please note that you must be in the directory containing the log_files subdirectory for this command to work.
sisana summarize
