CD_RF_microbiome

Scripts used for the processing and analysis of human genetic, metagenomic and 16S rRNA sequences from the biopsies of pediatric Crohn's patients and healthy colon controls.

Pre-processing

BISCUIT_utility_code.R contains useful functions and is sourced at the start of several of scripts in this repository.

raw_data/ contains the raw abundance tables used as input for the random forest analyses. biscuit_metadata.txt contains the relevant metadata for each sample.

These tables were prepped for the RF analyses with these scripts:

• prep_BISCUIT_16S_RF_input.R

• prep_BISCUIT_MGS_RF_input.R

• prep_combined_RF.R

Running RF

The RF models were run using this command-line R-script: run_RF.R - you can type this script with the -h option to see the input arguments.

The above script outputs RDS files for each RF model, which were parsed using commands in these two scripts:

get_summary_tables.R - Get summary metrics for each model.

get_varImp_tables.R - Get features sorted by variable importance for each model.

Several scripts contain the plotting commands used when analyzing this data and are found in plot_scripts/.

Human Genotype Parsing

Most of the human genotype analyses were performed with publicly available software. However, two scripts needed to be written to perform the imputation pipeline.

• splitVCFbyChr.pl - Split a VCF into different chromosomes.
• impute2vcf.pl - Parse the IMPUTE2 output and to re-build a VCF that includes the imputed variants.