From 616fd2bb4454a8d4d55daa0e88e5ec38d6e33dd3 Mon Sep 17 00:00:00 2001
From: Eric Coissac <eric.coissac@metabarcoding.org>
Date: Mon, 14 Feb 2022 15:10:47 +0100
Subject: [PATCH] A script for helping in clustering reference database for CDS
 annotation

Former-commit-id: 7babc60d47f433efd1301fbbe2a5714bfe7f7658
Former-commit-id: cf45c79769c6204598dd456573846496e4e834c0
---
 detectors/cds/tools/lib/clusterize_prot.sh | 73 ++++++++++++++++++++++
 1 file changed, 73 insertions(+)
 create mode 100644 detectors/cds/tools/lib/clusterize_prot.sh

diff --git a/detectors/cds/tools/lib/clusterize_prot.sh b/detectors/cds/tools/lib/clusterize_prot.sh
new file mode 100644
index 0000000..f14b2c1
--- /dev/null
+++ b/detectors/cds/tools/lib/clusterize_prot.sh
@@ -0,0 +1,73 @@
+
+
+CDHIT_ID=0.7
+CDHIT_DELTA=0.8
+
+
+function clusterize() {
+ 
+ 	local prot="${1}"
+ 	local fastain="${prot}.fst"
+ 	local cdhitout="${prot}.cdhit"
+ 	local fasta1="${prot}.1l.fst"
+ 	
+ 	rm -rf "${prot}"
+ 	mkdir -p "${prot}"
+ 	
+ 	cd-hit 	-i "${fastain}" \
+ 	    	-o "${prot}/${cdhitout}" \
+ 	    	-c ${CDHIT_DELTA} \
+ 	    	-G 1 \
+			-g 1 \
+			-aL 0.95 \
+			-s ${CDHIT_ID} \
+ 	    	-b 350 -p 1 \
+ 	    	-d 0 -n 3
+ 	       
+ 	fasta1line "${fastain}" > "${prot}/${fasta1}"
+ 	
+ 	pushd "$prot"
+ 	
+ 	rm -rf "*.cluster.*.ids"
+ 	$AwkCmd -v prot="$prot" \
+ 	        '/^>/   {cluster=$2} \
+ 	         ! /^>/ {sub("\\.\\.*$","",$3); \
+ 	                 filename=prot".cluster."cluster".ids"; \
+ 	                 print $3 >> filename ; \
+ 	                 close(filename) }' "${cdhitout}.clstr"
+ 	                
+ 	rm -f "../$prot.clean.fst"
+ 	                 
+ 	for ids in *.cluster.*.ids ; do
+ 		cluster=$(printf "%03d" $(echo "${ids}" | $AwkCmd -F'.' '{print $3}'))
+ 		size=$(wc -l "$ids" | $AwkCmd  '{print $1}')
+ 		fsize=$(printf '%05d' $size)
+
+ 		alignment="${prot}.cluster.${fsize}.${cluster}.fst"
+ 	    
+ 	    if (( size > 1 )) ; then
+	 	    egrep -f "$ids" -A 1 "${fasta1}" | \
+	 	    egrep -v -- '^--$' | \
+	 	    clustalo -i -  > "$alignment"
+	 	else
+	 	    egrep -f "$ids" -A 1 "${fasta1}" | \
+	 	    egrep -v -- '^--$' > "$alignment"
+	    fi
+	 		
+ 	    if (( size >= 5 )) ; then
+ 	    	egrep -f "$ids" -A 1 "${fasta1}" | \
+	 	    egrep -v -- '^--$' | \
+	 	    formatfasta >> "../$prot.clean.fst"
+ 	    fi
+ 	    
+ 	    rm -f "$ids"
+ 	    
+ 	done 
+ 	
+ 	rm -f "${fasta1}"
+ 	rm -f "${cdhitout}"
+ 	
+ 	popd
+ 	       
+}
+ 
\ No newline at end of file