rowling2392 / g2s

Deep learning predictor of stool microbiome configuration from oral microbiome data

g2s

Deep learning predictor of stool microbiome configuration from oral microbiome data

README

SPACE REQUIREMENTS : g2s needs 32 GB of free space and 16 GB of RAM memory to work on your device

OTHER REQUIREMENTS:

• R > version 3.5.0
• keras for R
• tensorflow for R

First steps

1. Download the g2s folder.

2. R settings and keras/tensorflow initialization

R
install.packages(keras)
library(keras)

Install tensorflow (It's only necessary to run this once.)

-> for GPU

install_keras(tensorflow = "gpu")

-> or CPU:

install_keras() 

quit()

Usage and Help

Rscript g2s.R otu_table_oral_microbiome.txt names_gingival_bacteria3.txt model_260719.h5 output_folder

• otu_table_oral_microbiome.txt: genus level (L6) relative abundance table with samples in the columns and the full taxonomy following the greengens 05_2013 style in the rows. Rel. Ab. must be 0 to 1. (INPUT)
• names_gengival_bacteria3.txt: this file is provided together with the g2s script and is necessary for automatically formatting the input file.
• model_260719.h5: this is the deep neural network implemented for doing the prediction
• output_folder: name of the output directory

Expected outputs

The tool provides two files within the output folder.

1. (.txt) Tabular outputs report the predicted structure of the stool microbiome in term of relative abundances.
2. (.pdf) Graphical representations (bar plots) of the predicted microbiome structures

Examples

For verifing the correct installation use the data within the test folder and compare the results you obtain with the files test_g2s.txt and test_g2s.pdf

Rscript g2s.R test/test.txt names_gingival_bacteria3.txt model_260719.h5 test_results