Enterotype-prediction-based-on-the-life-history

Machine learning classification pipe-line used for the study "Long-term life history predicts current elderly gut microbiome"

Hardware requirements

The pipe-line is designed to be run into Sun Grid Engine queuing cluster architecture (qsub script.sh), however, it can be run into a desktop computer but the for-loop implementation is extremely inefficient (Processor: Intel® Core™ i7-7820HQ CPU @ 2.90GHz × 8, Memory: 31 GiB, OS type: 64-bit)

Software requirements

R version 4.0.3 (2020-10-10)

R packages:

• caret_6.0-86
• DMwR_0.4.1
• ROSE_0.0-4
• pROC_1.17.0.1
• mltools_0.3.5

Run the pipeline

Ubuntu bash script

Run a single CV

nested_cross_validation_rf=./R/nested_cross_validation_rf.R
outer_cv_path="./CV_Data/All_years_Rum_CV_40_SCALE_TRUE.RData"
ent="Rum"  ### Enterotype
KfoldsInnerLoop=5 ### number of cv in the inner loop
outdir="All_years_Rum_CV_40_SCALE_TRUE"
index=38 # num form 1 to 40

Rscript --vanilla $nested_cross_validation_rf $outer_cv_path $ent $KfoldsInnerLoop $outdir $index # The script must be executed for all the 40 CV sub-datasets

Run into a for loop. IMPORTANT: The pipeline was designed to be run as an executable script into a batch server (qsub script.sh), the for loop example is extremely inefficient

filename='parameters.txt'
while read line; do
 Rscript --vanilla $nested_cross_validation_rf $line
done < $filename