annotate lissero/variant4b.py @ 11:40f397e29951 draft

Uploaded
author estrain
date Thu, 08 Feb 2024 20:40:49 +0000
parents 412d55f09755
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
10
412d55f09755 Uploaded
estrain
parents:
diff changeset
1 import sys
412d55f09755 Uploaded
estrain
parents:
diff changeset
2
11
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
3 def identify_variants_with_genes(input_file_path, output_file_path, simple_name):
10
412d55f09755 Uploaded
estrain
parents:
diff changeset
4 # Define the genes of interest
412d55f09755 Uploaded
estrain
parents:
diff changeset
5 genes_of_interest = ['LMO0737', 'ORF2110', 'ORF2819']
412d55f09755 Uploaded
estrain
parents:
diff changeset
6
412d55f09755 Uploaded
estrain
parents:
diff changeset
7 # Open the input file and read its lines
412d55f09755 Uploaded
estrain
parents:
diff changeset
8 with open(input_file_path, 'r') as file:
412d55f09755 Uploaded
estrain
parents:
diff changeset
9 lines = file.readlines()
412d55f09755 Uploaded
estrain
parents:
diff changeset
10
412d55f09755 Uploaded
estrain
parents:
diff changeset
11 # Check if the file has more than just the header
412d55f09755 Uploaded
estrain
parents:
diff changeset
12 if len(lines) <= 1:
412d55f09755 Uploaded
estrain
parents:
diff changeset
13 print("Input file does not contain enough data.")
412d55f09755 Uploaded
estrain
parents:
diff changeset
14 return
412d55f09755 Uploaded
estrain
parents:
diff changeset
15
412d55f09755 Uploaded
estrain
parents:
diff changeset
16 # Extract the column headers and find the indices of the genes of interest
412d55f09755 Uploaded
estrain
parents:
diff changeset
17 headers = lines[0].strip().split('\t')
412d55f09755 Uploaded
estrain
parents:
diff changeset
18 gene_indices = [headers.index(gene) for gene in genes_of_interest]
412d55f09755 Uploaded
estrain
parents:
diff changeset
19 serotype_index = headers.index('SEROTYPE')
412d55f09755 Uploaded
estrain
parents:
diff changeset
20
11
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
21 # Modify the header to include the new first column
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
22 modified_header = f"FileName\t{lines[0]}"
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
23
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
24 # Initialize a list to hold the modified lines, starting with the modified header
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
25 modified_lines = [modified_header]
10
412d55f09755 Uploaded
estrain
parents:
diff changeset
26
412d55f09755 Uploaded
estrain
parents:
diff changeset
27 # Process each data line in the input file
412d55f09755 Uploaded
estrain
parents:
diff changeset
28 for line in lines[1:]:
412d55f09755 Uploaded
estrain
parents:
diff changeset
29 data = line.strip().split('\t')
412d55f09755 Uploaded
estrain
parents:
diff changeset
30 # Check if the genes of interest are all present (marked as "FULL")
412d55f09755 Uploaded
estrain
parents:
diff changeset
31 if all(data[index] == 'FULL' for index in gene_indices):
412d55f09755 Uploaded
estrain
parents:
diff changeset
32 # Modify the SEROTYPE column to "4b variant"
412d55f09755 Uploaded
estrain
parents:
diff changeset
33 data[serotype_index] = "4b variant"
11
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
34 # Prepend the simple name to the line
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
35 modified_line = f"{simple_name}\t" + '\t'.join(data) + '\n'
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
36 # Add the modified line to the list
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
37 modified_lines.append(modified_line)
10
412d55f09755 Uploaded
estrain
parents:
diff changeset
38
412d55f09755 Uploaded
estrain
parents:
diff changeset
39 # Write the modified lines to the output file
412d55f09755 Uploaded
estrain
parents:
diff changeset
40 with open(output_file_path, 'w') as file:
412d55f09755 Uploaded
estrain
parents:
diff changeset
41 file.writelines(modified_lines)
412d55f09755 Uploaded
estrain
parents:
diff changeset
42
412d55f09755 Uploaded
estrain
parents:
diff changeset
43 print(f'Results written to {output_file_path}')
412d55f09755 Uploaded
estrain
parents:
diff changeset
44
412d55f09755 Uploaded
estrain
parents:
diff changeset
45 if __name__ == "__main__":
11
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
46 if len(sys.argv) != 4:
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
47 print("Usage: python script.py <input_file_path> <output_file_path> <simple_name>")
10
412d55f09755 Uploaded
estrain
parents:
diff changeset
48 sys.exit(1)
412d55f09755 Uploaded
estrain
parents:
diff changeset
49
412d55f09755 Uploaded
estrain
parents:
diff changeset
50 input_file_path = sys.argv[1]
412d55f09755 Uploaded
estrain
parents:
diff changeset
51 output_file_path = sys.argv[2]
11
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
52 simple_name = sys.argv[3]
40f397e29951 Uploaded
estrain
parents: 10
diff changeset
53 identify_variants_with_genes(input_file_path, output_file_path, simple_name)
10
412d55f09755 Uploaded
estrain
parents:
diff changeset
54