Hierarchical Meta-Storms (HMS) comprehensively calculates the dissimilarities of microbiome functional profiles by considering multi-level metabolic pathway hierarchy. It contains two core components: i) a dissimilarity algorithm that comprehensively calculates the distances among microbiome functional profiles by considering their multi-level metabolic pathway hierarchy among functional gene families, and ii) a PCoA implementation optimized by multi-thread parallel computing to rapidly parse out the beta-diversity pattern for thousands of samples. It takes the microbiome functional profiles of KO and their relative abundance as input, and computes and outputs their pairwise distance matrix and then the principle coordinates of PCoA.
In addition, we also encapsulate the C++ source codes as an R package (Github: hrms) using RcppArmadillo framework and submitted to CRAN, making the kernel functions of calculating the microbiome distances and principle coordinates could be callable by R interpreter in both R terminal and R scripts.
Hierarchical Meta-Storms only requires a standard computer with sufficient RAM to support the operations defined by a user. For typical users, this would be a computer with about 2 GB of RAM. For optimal performance, we recommend a computer with the following specs:
RAM: 8+ GB
CPU: 4+ cores, 3.3+ GHz/core
OpenMP library is the C/C++ parallel computing library. Most Linux releases have OpenMP already been installed in the system. In Mac OS X, to install the compiler that supports OpenMP, we recommend using the Homebrew package manager:
brew install gcc
At present, Hierarchical Meta-Storms provides a fully automatic installer for easy installation.
a. Download the package
git clone https://github.com/qdu-bioinfo/hierarchical-meta-storms.git
b. Install by installer
cd hierarchical-meta-storms
source install.sh
The package should take less than 1 minute to install on a computer with the specifications recommended above.
The example dataset could be found at “example” folder. Check the “example/Readme” for details about the demo run.
If the automatic installer fails, Hierarchical Meta-Storms can still be installed manually.
a. Download the package
git clone https://github.com/qibebt-bioinfo/hierarchical-meta-storms.git
b. Configure the environment variables (the default environment variable configuration file is “~/.bashrc”)
export HierarchicalMetaStorms=Path to Hierarchical Meta-Storms
export PATH=”$PATH:$HierarchicalMetaStorms/bin/”
source ~/.bashrc
c. Compile the source code
cd hierarchical-meta-storms
make
a. Compute the distance matrix
MS-comp-func -T samples.ko.table -o samples.func.dist
The output file “samples.func.dist” is the pairwise distance matrix.
b. Implement the PCoA
MS-get-pc -d samples.func.dist -o samples.pcoa.pc
The output file “samples.pcoa.pc” is the coordinates results of PCoA.
Here we provide a demo dataset (Synthetic Dataset 1) with functional KO profiles of 30 artificial microbiomes in “example” folder. In this package, “dataset1.ko.abd” is the relative abundance on KOs-level, and “dataset1.meta” is the group information of the samples.
To run the demo, you can either:
cd example
sh Readme
or type the following command:
MS-comp-func -T dataset1.ko.abd -o dataset1.dist
MS-get-pc -d dataset1.dist -m dataset1.meta -o dataset1.pcoa.pc
Then the output file “dataset1.dist” is the pairwise distance of the 30 samples and "dataset1.pcoa.pc" is the coordinates results of PCoA based on the distance matrix.
This demo run should take less than 1 minute on a recommended computer.
a. MS-comp-func
It calculates the hierarchical meta-storms distance matrix among microbiome functional profiles. Run:
MS-comp-func -h
for detailed parameters.
b. MS-get-pc
It calculates the PCoA based the distance matrix. Run:
MS-get-pc -h
for detailed parameters.
Synthetic Dataset 1 contains 30 synthetic samples.
Synthetic Dataset 2 contains 30 synthetic samples.
Real Dataset 1 contains 20,000 real samples that are inferred from 16S rRNA gene by PICRUSt 2.
Real Dataset 2 contains 5,350 real samples that are inferred from 16S rRNA gene by PICRUSt 2.
Real Dataset 3 contains 2,354 real samples, reconstructed from WGS data by HUMANn 2.