mgCST-classifier

Requirements

R version 4.2.2 (2022-10-31) or higher
randomForestSRC version 3.2.1
gplots veresion 3.1.3

To Use:

When used in concert with the vaginal non-redundant gene database (VIRGO), assigns metagenomic community state types (mgCSTs) to vaginal metagenomes. This script requires VIRGO output files "summary.Abundance.txt" and "summary.NR.abundance.txt" as well as VIRGO reference file 0.geneLength.txt.

1. Clone this repository
2. Run data through VIRGO (https://virgo.igs.umaryland.edu).
3. Download the metagenomic subspecies random forest classifier from https://figshare.com/ndownloader/files/36123449 and place in the github repository directory.
4. Run classifier. The classifier will first assign metagenomic subspecies using the mgss_classifier.RDS file downloaded from figshare. It will then assign metagenomic CSTs to each sample using a nearest-neighbor, centroid-based classifier.

Full paths to the following files/directories must be supplied in this order:
(1) summary.Abundance.txt
(2) summary.NR.abundance.txt
(3) VIRGO-master directory
(4) mgCST-classifier-master directory

Example:

Rscript classify_mgCST_centroid.R example/summary.Abundance.txt example/summary.NR.abundance.txt /full/path/to/VIRGO-master /full/path/to/mgCST-classifier-master

Output is written to current directory. Each output file is dated.

Output files:

relabund_w_mgCSTs_1Jul2022.csv: Relative abundances of metagenomic subspecies (columns) by samples (rows). Right-most column contains mgCST classification.

mgCST_heatmap_1Jul2022.pdf: A heatmap of relative abundances of 50 most abundant metagenomic subspecies with samples colored by mgCST.

mgCSTs_1Jul2022.csv: Two-column table with sample ID and mgCST assignment

norm_counts_mgSs_mgCST_1Jul2022.csv: Gene abundances normalized by gene length of metagenomic subspecies (columns) by samples (rows). Right-most column contains mgCST classification.

NOTE: Gene Count Normalization assumes 150 bp long reads: (Gene_Abundance*150)/Gene_Length