While there have been substantial advances in understanding microbial virulence in cultured systems, a metagenomic examination of microbial virulence can capture community relationships and environmental context necessary for understanding disease progression. However, setting an appropriate problem scope for identifying new virulence factors can be challenging as certain contexts, such as an infectious process, can significantly alter the landscape of the microbiome and the pathological potential of individual microbes, including those normally considered commensal. Crucially, few tools exist for probing microbial virulence in metagenomes specifically, constraining the development of culture-free methods in public health and epidemiology.
We used parallel machine learning methods to approach the problem of characterizing virulence factors (VF) in diseased and healthy metagenomes. Using genes from the core set of the Virulence Factor Database (http://www.mgc.ac.cn/VFs/), we used an HMM to profile known virulence factors and apply profiles to diseased and healthy metagenomes. In parallel to this approach, we used a set of labelled pathological and commensal genomes and subtracted the VFDB virulence factor genes from both sets. We then trained the VF-subtracted genomes on an SVM model to classify pathogenic and non-pathogenic genomes. Both techniques form a complementary approach to VF characterization by using well-characterized virulence factors and commensal genomes to profile similar characteristics in the metagenome space (HMM), and by exploring the potential for virulent genes uncharacterized by the VF dataset within the same metagenomes. This combination of techniques can provide spatially-resolved scoring within the metagenome to identify potential virulence factors.
Currently the VirFac repo is provided as a Python project that can be installed via setuptools. It requires Python 3.7 or later and some package dependencies that can be installed via pip:
pip install -r requirements.txt
requests==2.22.0
click==7.0
numpy==1.13.3
pandas=0.25.0
scikit-learn==0.21.3
Eventually the project will be available on PyPI.
Experimentally verified virulence factor genes from the Virulence Factor Database (VFDB: Chen et al 2015, Accessed 8/13/19 https://academic.oup.com/nar/article/44/D1/D694/2503049) were used to represent virulence-associated genes. Example metagenomes used for testing were drawn from public datasets listed on NCBI SRA and included healthy and disease-state human skin metagenomic samples. Specifically, diseased metagenomes were drawn from the Diabetic Foot Ulcer metagenome study (BioProject: PRJNA506988) and healthy foot skin metagenomes were drawn from BioProject: PRJEB30094. Metagenomes were assembled using MetaSPADES. (Nurk et al https://www.ncbi.nlm.nih.gov/pubmed/28298430 )
A Hidden Markov Model (HMM) was applied to the VFDB genes to create virulence profiles. Genes were selected for which at least five different bacterial species were available. Multiple sequence alignments were generated using MUSCLE [1] and HMMs using HMMER3 [2]. Genomes and/or corresponding protein coding sequences were screened with HMMSEARCH[2] using pre-computed significance scores. Scores were calculated as 80% of the envelope alignment score of a representative sequence corresponding to its HMM. Alignments were filtered by custom scripts to extract putative virulence factors’ loci. VF sequences were concatenated, aligned and used as input for phylogenetic analyses. Phylogenetic trees were constructed using RAXML-ng [3] and analyzed using R package Ape[4] and Newick Utilities[5]. Virulence tags were assigned based on the number of virulence loci found and phylogenetic classification. All analyses are described in Snakemake pipeline[6].
A SVM model was also developed to classify virulent and non-virulent gene segments by training on a reference set of labelled pathogen and commensal genomes. The pathogen genomes were acquired from an NCBI Assembly search and included the species identified in the VFDB dataset. Commensal genomes were also acquired from an NCBI Assembly search, and included species selected from the NHSN Common Commensals List (https://www.cdc.gov/nhsn/pdfs/pscmanual/4psc_clabscurrent.pdf) and from Busby et al 2012 (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5866053/).
Data is available from the NCBI Sequence Read Archive under projects ERP112507 and SRP170931. "VirFac" software is made available under the MIT License (see LICENSE.)
Ousmane H. Cissé
[email protected]
NIH Clinical Center
Bethesda MD, 20814
Shennan Lu
Liz Norred
[email protected]
Bredesen Center for Interdisciplinary Research and Education, University of Tennessee, Knoxville, TN 37922
Knoxville TN, 37922
Justin Payne
[email protected]
US-FDA Center for Food Safety and Applied Nutrition
College Park MD, 20710
Sherry Bhalla
[email protected]
Icahn School of Medicine at Mount Sinai
New York, NY, 20019
Diabetic Foot Ulcer dataset provided courtesy of UMaryland/CosmosID and described at
https://www.ncbi.nlm.nih.gov/bioproject/PRJNA506988
Healthy and diabetic foot microbiomes provided courtesy of CLM and described at
https://www.ncbi.nlm.nih.gov/bioproject/PRJEB30094
This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.
Cookiecutter: https://github.com/audreyr/cookiecutter
https://github.com/audreyr/cookiecutter-pypackage
This was a Microbial Virulence in the Cloud Hackathon 2019 project.