#!/bin/bash
# Script name: run_interproscan.sh
# Function: Batch process .faa files with InterProScan
# Usage: ./run_interproscan.sh <input_directory> <output_directory>
# Example: ./run_interproscan.sh ./input_genes ./ph_predict

# Check input parameters (must be 2: input directory and output directory)
if [ $# -ne 2 ]; then
    echo "Error: Incorrect number of arguments!"
    echo "Usage: $0 <input_directory> <output_directory>"
    echo "Example: $0 ./prokka_annotation ./ph_predict"
    exit 1
fi

input_dir="$1"
output_root="$2"  # Output directory from parameter

# Create main output directory
echo "Creating main output directory: $output_root"
mkdir -p "$output_root" || {
    echo "Error: Failed to create output directory $output_root!"
    exit 1
}

# Find all .faa files in input directory (including subdirectories)
echo "Searching for .faa files in $input_dir..."
FAA_FILES=$(find "$input_dir" -type f -name "*.faa")

# Check if any .faa files were found
if [ -z "$FAA_FILES" ]; then
    echo "Error: No .faa files found in $input_dir (including subdirectories)!"
    exit 1
fi

# Copy all .faa files to output directory
echo "Copying .faa files to $output_root..."
cp -v $FAA_FILES "$output_root/" || {
    echo "Error: Failed to copy .faa files to $output_root!"
    exit 1
}

# Process each .faa file with InterProScan
echo "Starting InterProScan analysis..."
for faa_file in "$output_root"/*.faa; do
    # Skip if not a valid file (e.g., empty glob)
    [ -f "$faa_file" ] || continue

    # Extract sample name (remove path and .faa suffix)
    sample_name=$(basename "$faa_file" .faa)
    
    # Define output files for this sample
    # ips_output="${output_root}/${sample_name}_interproscan.tsv"  # InterProScan result
    # filtered_output="${output_root}/${sample_name}_filtered.tsv"  # Filtered result
    # gene_list="${output_root}/${sample_name}_gene_list.txt"        # Extracted gene names

    echo "Processing $faa_file..."
    # Run InterProScan (adjust parameters as needed; example uses common options)
    /media/interproscan-5.75-106.0/interproscan.sh \
        -i "$faa_file" \
        -d "$output_root" \
        -f tsv \
        -cpu 8 \
        -appl Pfam \
        > "${output_root}/${sample_name}_ips.log" 2>&1

    # Check if InterProScan succeeded
    if [ $? -ne 0 ]; then
        echo "Warning: InterProScan failed for $faa_file (see log: ${output_root}/${sample_name}_ips.log)"
        continue
    fi

    # Filter results (example: retain rows with significance; adjust columns/conditions as needed)
    # awk -F'\t' 'NR == 1 || $8 < 1e-5' "$ips_output" > "$filtered_output"

    # Extract gene names (column 1 in InterProScan TSV)
    # awk -F'\t' 'NR > 1 {print $1}' "$filtered_output" | sort -u > "$gene_list"

    echo "Completed processing $sample_name:"
    echo "  Raw results: $ips_output"
    # echo "  Filtered results: $filtered_output"
    # echo "  Gene list: $gene_list"
done

echo "All InterProScan analyses completed! Results in $output_root"