Added predictive modeling capabilities

main.sh

@@ -164,7 +164,7 @@ qiime composition add-pseudocount \
  --i-table feature-table.qza \
  --o-composition-table composition-table.qza
 
-# Run ancom for CowID, Age, TrmtGroup
+# Run ancom for all categories in catcols
 # Once again, QIIME only uses one processor (even though this
 # is a HUGE task), so we should parallelize it for speed
 cat catcols.txt | \
@@ -175,6 +175,16 @@ cat catcols.txt | \
    --m-metadata-column {} \
    --o-visualization "visualizations/ancom-{}.qzv" \
    --verbose
-echo "--^-- X: Performing ANCOM...Done!"	 
+echo "--^-- X: Performing ANCOM...Done!"
+
+# Create category-based predictive models
+cat catcols.txt | \
+  xargs -P"$SLURM_NTASKS" -L1 srun -n1 -N1 --exclusive \
+  ./sample-classifier.sh
+
+# Create continuous predictive models
+cat numcols.txt | \
+  xargs -P"$SLURM_NTASKS" -L1 srun -n1 -N1 --exclusive \
+  ./sample-regression.sh
 
 echo "All Done!"

sample-classifier.sh

@@ -0,0 +1,49 @@
+#!/bin/bash
+#SBATCH --account=cowusda2016
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=4
+#SBATCH --mem=8G
+
+# Fetch the category we're working with from upstream
+CATEGORY=${1%}
+
+# Save the folder name we will be saving everything into
+FOLDERNAME="${CATEGORY}-classifier"
+
+# Load the required modules
+module restore system
+module load swset
+module load miniconda3
+
+# Start up qiime
+source activate qiime2
+
+# Make sure we have a clean slate to work with
+echo "--^-- X: Clearing previous classifier results..."
+rm -r "$FOLDERNAME"
+echo "--^-- X: Clearing previous classifier results...Done!"
+
+# Solve the model
+echo "--^-- X: Constructing model..."
+qiime sample-classifier classify-samples \
+ --i-table feature-table.qza \
+ --m-metadata-file metadata.tsv \
+ --m-metadata-column "$CATEGORY" \
+ --p-n-jobs 4 \
+ --p-missing-samples ignore \
+ --p-optimize-feature-selection \
+ --output-dir "$FOLDERNAME" \
+ --verbose
+echo "--^-- X: Constructing model...Done!"
+
+# Convert the model output into readable visualizations
+echo "--^-- X: Making visualizations..."
+qiime metadata tabulate \
+ --m-input-file "${FOLDERNAME}/feature_importance.qza" \
+ --o-visualization "${FOLDERNAME}/feature-importance.qzv"
+
+qiime metadata tabulate \
+ --m-input-file "${FOLDERNAME}/predictions.qza" \
+ --m-input-file metadata.tsv \
+ --o-visualization "${FOLDERNAME}/predictions.qzv"
+echo "--^-- X: Making visualizations...Done!"

sample-regression.sh

@@ -0,0 +1,49 @@
+#!/bin/bash
+#SBATCH --account=cowusda2016
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=4
+#SBATCH --mem=8G
+
+# Fetch the category we're working with from upstream
+CATEGORY=${1%}
+
+# Save the folder name we will be saving everything into
+FOLDERNAME="${CATEGORY}-regression"
+
+# Load the required modules
+module restore system
+module load swset
+module load miniconda3
+
+# Start up qiime
+source activate qiime2
+
+# Make sure we have a clean slate to work with
+echo "--^-- X: Clearing previous regression results..."
+rm -r "$FOLDERNAME"
+echo "--^-- X: Clearing previous regression results...Done!"
+
+# Solve the model
+echo "--^-- X: Constructing model..."
+qiime sample-classifier regress-samples \
+ --i-table feature-table.qza \
+ --m-metadata-file metadata.tsv \
+ --m-metadata-column "$CATEGORY" \
+ --p-n-jobs 4 \
+ --p-missing-samples ignore \
+ --p-optimize-feature-selection \
+ --output-dir "$FOLDERNAME" \
+ --verbose
+echo "--^-- X: Constructing model...Done!"
+
+# Convert the model output into readable visualizations
+echo "--^-- X: Making visualizations..."
+qiime metadata tabulate \
+ --m-input-file "${FOLDERNAME}/feature_importance.qza" \
+ --o-visualization "${FOLDERNAME}/feature-importance.qzv"
+
+qiime metadata tabulate \
+ --m-input-file "${FOLDERNAME}/predictions.qza" \
+ --m-input-file metadata.tsv \
+ --o-visualization "${FOLDERNAME}/predictions.qzv"
+echo "--^-- X: Making visualizations...Done!"