MetagenomicKG is a novel metagenomics knowledge graph that integrates commonly used taxonomic information (GTDB), functional annotations (KEGG), pathogenicity resources (BV-BRC), and other biomedical knowledge.
The associated preprint can be found at: https://www.biorxiv.org/content/10.1101/2024.03.14.585056v1. Please cite via:
Ma, C., Liu, S., & Koslicki, D. (2024). MetagenomicKG: A Knowledge Graph for Metagenomic Applications. bioRxiv, 2024-03.
- Metagenomic
MetagenomicKG integrates knowledge from 7 relevant data sources: GTDB taxonomy, NCBI taxonomy, KEGG, RTX-KG2, BV-BRC, MicroPhenoDB, and NCBI AMRFinderPlus Prediction. It consists of 14 node types and 33 edge types (see statistics in the table below).
We have provided the pre-built version of MetagenomicKG, you can download the MetagenomicKG.zip file from Zenodo. We also host a neo4j instance for MetagenomicKG: http://mkg.cse.psu.edu:7474/ (Username:neo4j, Password:klabneo4j). If you would like to rebuild it or reproduce the use case reulsts reported in our paper, you can follow the instruction below.
Node Statistics
| Node Type | Node Count |
|---|---|
| Microbe | 95,2191 |
| Disease | 10,3129 |
| Phenotypic Feature | 88,708 |
| KO | 26,588 |
| Compound | 19,265 |
| Reaction | 15,208 |
| Drug | 12,347 |
| Glycan | 11,225 |
| Enzyme | 8,109 |
| AMR | 5,147 |
| Drug Group | 2,458 |
| Network | 1,525 |
| Pathway | 569 |
| Module | 481 |
| 1,246,950 |
Edge Statistics
| Edge Type | Edge Count |
|---|---|
| genetically associated with | 42,134,426 |
| associated with | 10,802,017 |
| subclass of | 1,141,279 |
| superclass of | 981,435 |
| physically interacts with | 458,541 |
| has participant | 192,830 |
| participates in | 189,781 |
| related to | 95,286 |
| chemically similar to | 56,055 |
| biomarker for | 47,698 |
| has phenotype | 30,291 |
| treats | 21,049 |
| close match | 17,389 |
| has part | 13,981 |
| catalyzes | 8,819 |
| contraindicated for | 5,158 |
| same as | 3,887 |
| is sequence variant of | 2,026 |
| gene product of | 2,026 |
| correlated with | 1,619 |
| contributes to | 286 |
| temporally related to | 252 |
| affects | 176 |
| causes | 171 |
| has input | 82 |
| produces | 27 |
| has metabolite | 23 |
| actively involved in | 21 |
| gene associated with condition | 9 |
| disrupts | 4 |
| derives from | 1 |
| disease has basis in | 1 |
| located in | 1 |
| 56,206,647 |
We recommend using a virtual environment to ensure a clean and isolated workspace for reproducibility. This can be accomplished using either Conda or Mamba (a faster alternative to Conda).
To create your Conda environment, follow these steps:
# Clone the MetagenomicKG repository
git clone https://github.com/KoslickiLab/MetagenomicKG.git
cd MetagenomicKG
# Please note that the versioin of pytorch we used might not be compatible with your nvidia cuda version. So, please first check your version and change it in metagenomickg_env.yml if needed.
conda env create -f envs/metagenomickg_env.yml
# Download required files for the package 'pytaxonkit'
wget -c ftp://ftp.ncbi.nih.gov/pub/taxonomy/taxdump.tar.gz
tar -zxvf taxdump.tar.gz
mkdir -p $HOME/.taxonkit
cp names.dmp nodes.dmp delnodes.dmp merged.dmp $HOME/.taxonkit
# Activate the newly created environment
conda activate metagenomickg_envIf you prefer using Mamba instead of Conda, just simply repalce conda with mamba in the above commands.
Before rebuilding MetagenomicKG and replicating the results of use cases, you can should modify some global variables in the config.yaml file. We listed some required variables below:
Since we utilize the Unified Medical Language System (UMLS) search function via UMLS APIs for identifier mapping, you should first get a UMLS API key (please follow this instruction to get one). After that, replace UMLS_API_KEY with your API key in the config.yaml file.
MetagenomicKG includes KEGG data downloaded from KEGG FTP. According to KEGG policy, we can provide this dataset. To obtain this dataset, you can follow this instruction. Once you download data, you should replace the path of KEGG_FTP_DATA_DIR key in the config.yaml file.
We constructed an automatic pipeline to rebuild MetagenomicKG via Snakemake. Since MetagenomicKG uses RTX-KG2, which includes UMLS data, you need to contact authors to demonstrate that you have accepted the license terms in order to get access to download KG2. Once you have access, please download and put the kg2c-tsv.tar.gz file to ./data/RTX_KG2 folder.
After downloading the RTX-KG2 TSV files is done, you can run the pipeline via:
snakemake --cores 16 -s run_buildKG_pipeline.smk targetsOnce it is completed, you can find merged node and edge TSV files (KG_nodes_v6.tsv and KG_edges_v6.tsv) from ./data/merged_KG folder.
Once you login the neo4j instance of MetagenomicKG (see login info in About Graph Section), you can use the following Neo4j Cypher Query to replicate what we show in the paper. In this query, we find at most 10 paths of protein - pathogen with name 'Staphylococcus aureus' - KO - pathway - disease - drug.
MATCH p=(n0:`biolink:Protein`)-[]-(n1:`biolink:OrganismTaxon`)-[]-(n2:`biolink:BiologicalEntity`)-[]-(n3:`biolink:Pathway`)-[]-(n4:`biolink:Disease`)-[]-(n5:`biolink:Drug`)
WHERE n1.is_pathogen = "True" and ANY (n1_names IN n1.all_names WHERE n1_names contains 'Staphylococcus aureus')
RETURN p LIMIT 10
To replicate the results of use case 2, you can simply run the Snakemake pipelie via:
snakemake --cores 16 -s run_usecase2_pipeline.smk targetsTo replicate the results of use case 3, you can simply run the Snakemake pipelie via:
snakemake --cores 16 -s run_usecase3_pipeline.smk targetsIf you have any questions or need help, please contact @chunyuma or @ShaopengLiu1 or @dkoslicki.