Merge pull request #33 from YuWei-CH/support-new-chromosome

mohammedkhalfan · web-flow · commit 880677899194 · 2025-05-23T22:12:02.000-04:00
Support add new chromosome sequence feature
diff --git a/reform.py b/reform.py
diff --git a/test_data/17/gold.fa b/test_data/17/gold.fa
@@ -0,0 +1,4 @@
+>X
+ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK
+>Y
+AAAATTTTGGGGCCCC
diff --git a/test_data/17/gold.gtf b/test_data/17/gold.gtf
@@ -0,0 +1,8 @@
+X	ref	exon	5	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	CDS	8	22	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	start_codon	5	7	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	stop_codon	23	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+Y	ref	exon	1	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	CDS	4	14	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	start_codon	1	3	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	stop_codon	14	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
diff --git a/test_data/17/in.fa b/test_data/17/in.fa
@@ -0,0 +1,2 @@
+>Y
+AAAATTTTGGGGCCCC
diff --git a/test_data/17/in.gtf b/test_data/17/in.gtf
@@ -0,0 +1,4 @@
+Y	ref	exon	1	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	CDS	4	14	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	start_codon	1	3	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	stop_codon	14	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
diff --git a/test_data/17/ref.fa b/test_data/17/ref.fa
@@ -0,0 +1,2 @@
+>X
+ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK
diff --git a/test_data/17/ref.gtf b/test_data/17/ref.gtf
@@ -0,0 +1,4 @@
+X	ref	exon	5	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	CDS	8	22	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	start_codon	5	7	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	stop_codon	23	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
diff --git a/test_data/18/gold.fa b/test_data/18/gold.fa
@@ -0,0 +1,8 @@
+>X
+ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK
+>Y
+AAAATTTTGGGGCCCC
+>H
+GGGGAATTCCCCGGGG
+>M
+CCCCGGGGAAAATTTT
diff --git a/test_data/18/gold.gtf b/test_data/18/gold.gtf
@@ -0,0 +1,18 @@
+X	ref	exon	5	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	CDS	8	22	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	start_codon	5	7	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	stop_codon	23	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+##sequence-region Y 1 16
+Y	ref	exon	1	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	CDS	4	14	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	start_codon	1	3	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	stop_codon	14	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+##sequence-region H 1 16
+H	ref	exon	1	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	CDS	4	14	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	start_codon	1	3	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	stop_codon	14	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+M	ref	exon	1	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	CDS	4	14	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	start_codon	1	3	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	stop_codon	14	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
diff --git a/test_data/18/in1.fa b/test_data/18/in1.fa
@@ -0,0 +1,2 @@
+>Y
+AAAATTTTGGGGCCCC
diff --git a/test_data/18/in1.gtf b/test_data/18/in1.gtf
@@ -0,0 +1,5 @@
+##sequence-region Y.11 1 16
+Y	ref	exon	1	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	CDS	4	14	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	start_codon	1	3	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	stop_codon	14	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
diff --git a/test_data/18/in2.fa b/test_data/18/in2.fa
@@ -0,0 +1,2 @@
+>H
+GGGGAATTCCCCGGGG
diff --git a/test_data/18/in2.gtf b/test_data/18/in2.gtf
@@ -0,0 +1,5 @@
+##sequence-region F.11 3 27
+H	ref	exon	1	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	CDS	4	14	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	start_codon	1	3	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	stop_codon	14	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
diff --git a/test_data/18/in3.fa b/test_data/18/in3.fa
@@ -0,0 +1,2 @@
+>M
+CCCCGGGGAAAATTTT
diff --git a/test_data/18/in3.gtf b/test_data/18/in3.gtf
@@ -0,0 +1,4 @@
+M	ref	exon	1	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	CDS	4	14	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	start_codon	1	3	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	stop_codon	14	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
diff --git a/test_data/18/ref.fa b/test_data/18/ref.fa
@@ -0,0 +1,2 @@
+>X
+ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK
diff --git a/test_data/18/ref.gtf b/test_data/18/ref.gtf
@@ -0,0 +1,4 @@
+X	ref	exon	5	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	CDS	8	22	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	start_codon	5	7	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	stop_codon	23	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
diff --git a/test_data/19/gold.fa b/test_data/19/gold.fa
@@ -0,0 +1,8 @@
+>X
+ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK
+>Y
+AAAATTTTGGGGCCCC
+>H
+GGGGAATTCCCCGGGG
+>M
+CCCCGGGGAAAATTTT
diff --git a/test_data/19/gold.gtf b/test_data/19/gold.gtf
@@ -0,0 +1,18 @@
+X	ref	exon	5	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	CDS	8	22	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	start_codon	5	7	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	stop_codon	23	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+##sequence-region Y 1 16
+Y	ref	exon	1	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	CDS	4	14	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	start_codon	1	3	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Y	ref	stop_codon	14	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+##sequence-region H 1 16
+H	ref	exon	1	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	CDS	4	14	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	start_codon	1	3	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	stop_codon	14	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+M	ref	exon	1	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	CDS	4	14	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	start_codon	1	3	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	stop_codon	14	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
diff --git a/test_data/19/in1.fa b/test_data/19/in1.fa
@@ -0,0 +1,2 @@
+>Z
+AAAATTTTGGGGCCCC
diff --git a/test_data/19/in1.gtf b/test_data/19/in1.gtf
@@ -0,0 +1,5 @@
+##sequence-region Y.11 1 16
+Z	ref	exon	1	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Z	ref	CDS	4	14	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Z	ref	start_codon	1	3	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
+Z	ref	stop_codon	14	16	.	+	0	gene_id "new_gene"; transcript_id "new_gene.1";
diff --git a/test_data/19/in2.fa b/test_data/19/in2.fa
@@ -0,0 +1,2 @@
+>Y
+GGGGAATTCCCCGGGG
diff --git a/test_data/19/in2.gtf b/test_data/19/in2.gtf
@@ -0,0 +1,5 @@
+##sequence-region F.11 3 27
+H	ref	exon	1	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	CDS	4	14	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	start_codon	1	3	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
+H	ref	stop_codon	14	16	.	+	0	gene_id "gene2"; transcript_id "gene2.1";
diff --git a/test_data/19/in3.fa b/test_data/19/in3.fa
@@ -0,0 +1,2 @@
+>M
+CCCCGGGGAAAATTTT
diff --git a/test_data/19/in3.gtf b/test_data/19/in3.gtf
@@ -0,0 +1,5 @@
+##Test Data
+M	ref	exon	1	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+K	ref	CDS	4	14	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+Z	ref	start_codon	1	3	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
+M	ref	stop_codon	14	16	.	+	0	gene_id "gene4"; transcript_id "gene4.1";
diff --git a/test_data/19/ref.fa b/test_data/19/ref.fa
@@ -0,0 +1,2 @@
+>X
+ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK
diff --git a/test_data/19/ref.gtf b/test_data/19/ref.gtf
@@ -0,0 +1,4 @@
+X	ref	exon	5	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	CDS	8	22	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	start_codon	5	7	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
+X	ref	stop_codon	23	25	.	+	0	gene_id "ref_gene"; transcript_id "ref_gene.1";
diff --git a/test_reform.py b/test_reform.py
@@ -684,5 +684,123 @@ def test_case_16(self):
 
 		os.chdir(wd)
 
+	def test_case_17(self):
+		"""
+		Case 17:
+		Testing Reform which adding new chrom
+		"""
+
+		wd = os.getcwd()
+		os.chdir('test_data/17/')
+
+		command = """
+		python3 ../../reform.py \
+		--new_chrom="Y" \
+		--in_fasta=in.fa \
+		--in_gff=in.gtf \
+		--ref_fasta=ref.fa \
+		--ref_gff=ref.gtf
+		"""
+
+		response = subprocess.getoutput(command)
+		print(response)
+
+		with open('gold.gtf', 'r') as f:
+			gold_gff = f.read()
+		with open('ref_reformed.gtf', 'r') as f:
+			new_gff = f.read()
+		print("Testing gtf")
+		self.assertListEqual(list(gold_gff), list(new_gff))
+		print("Done")
+
+		with open('gold.fa', 'r') as f:
+			gold_fa = f.read()
+		with open('ref_reformed.fa', 'r') as f:
+			new_fa = f.read()
+		print("Testing Fasta")
+		self.assertListEqual(list(gold_fa), list(new_fa))
+		print("Done")
+
+		os.chdir(wd)
+	
+	def test_case_18(self):
+		"""
+		Case 18:
+		Testing Reform which adding multiple new chroms
+		"""
+
+		wd = os.getcwd()
+		os.chdir('test_data/18/')
+
+		command = """
+		python3 ../../reform.py \
+		--new_chrom="Y,H,M" \
+		--in_fasta=in1.fa,in2.fa,in3.fa \
+		--in_gff=in1.gtf,in2.gtf,in3.gtf \
+		--ref_fasta=ref.fa \
+		--ref_gff=ref.gtf
+		"""
+
+		response = subprocess.getoutput(command)
+		print(response)
+
+		with open('gold.gtf', 'r') as f:
+			gold_gff = f.read()
+		with open('ref_reformed.gtf', 'r') as f:
+			new_gff = f.read()
+		print("Testing gtf")
+		self.assertListEqual(list(gold_gff), list(new_gff))
+		print("Done")
+
+		with open('gold.fa', 'r') as f:
+			gold_fa = f.read()
+		with open('ref_reformed.fa', 'r') as f:
+			new_fa = f.read()
+		print("Testing Fasta")
+		self.assertListEqual(list(gold_fa), list(new_fa))
+		print("Done")
+
+		os.chdir(wd)
+
+	def test_case_19(self):
+		"""
+		Case 19:
+		Testing Reform with incorrect new chrom and comments in input FASTA 
+  		and annotation files. Also, testing reform with more formal printing
+		"""
+
+		wd = os.getcwd()
+		os.chdir('test_data/19/')
+
+		command = """
+		python3 ../../reform.py \
+		--new_chrom="Y,H,M" \
+		--in_fasta=in1.fa,in2.fa,in3.fa \
+		--in_gff=in1.gtf,in2.gtf,in3.gtf \
+		--ref_fasta=ref.fa \
+		--ref_gff=ref.gtf
+		"""
+
+		response = subprocess.getoutput(command)
+		print(response)
+
+		with open('gold.gtf', 'r') as f:
+			gold_gff = f.read()
+		with open('ref_reformed.gtf', 'r') as f:
+			new_gff = f.read()
+		print("Testing gtf")
+		self.assertListEqual(list(gold_gff), list(new_gff))
+		print("Done")
+
+		with open('gold.fa', 'r') as f:
+			gold_fa = f.read()
+		with open('ref_reformed.fa', 'r') as f:
+			new_fa = f.read()
+		print("Testing Fasta")
+		self.assertListEqual(list(gold_fa), list(new_fa))
+		print("Done")
+
+		os.chdir(wd)
+
 if __name__ == '__main__':
     unittest.main()

-Original file line number
+Diff line change
@@ @@ -0,0 +1,4 @@ @@
 +>X
 +ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK
 +>Y
 +AAAATTTTGGGGCCCC
Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	`+>X`
	`2`	`+ZZZZABBBBBDDDDDCCCCCIIIIIKKKKK`