我们都知道，利用R包infercnv对scRNA-seq数据进行CNV推断时，首个步骤是运行CreateInfercnvObject()函数构建infercnv对象，此处必须设置gene_order_file参数，其输入是一个基因的染色体位置信息文件，以制表符分隔。
inferCNV作为TrinityCTAT Toolkit的一个组成部分，一些版本的 Genomic Position Files 已经生成过并且放置在 https://data.broadinstitute.org/Trinity/CTAT/cnv/ 供大家获取。
可以看出数据的版本比较老，有些基因组注释文件还是依赖hg19参考基因组，而我们现在表达定量，特别是10x数据，上游一般直接用Cell Ranger流程，官网目前给出的集成好的参考基因组相关内容的压缩包refdata-gex-GRCh38-2020-A.tar.gz内的文件都是基于hg38。如果还用TrinityCTAT给出的老数据作为InferCNV gene_order_file参数的输入，得到的结果总会是相对粗糙的。那如果我们想自己构建一个 Genomic Position File 呢？贴心的 developer 造了一个现成的轮子，我们把轮子下载下来。同时下载与10x官网给出信息一致对应的gencode.v32文件。
cd /mnt/d/Bioinfo/Single_Cell/inferCNV/
wget -c https://github.com/broadinstitute/infercnv/raw/master/scripts/gtf_to_position_file.py
wget -c http://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_32/gencode.v32.primary_assembly.annotation.gtf.gz
运行时报错，443端口raw.githubusercontent.com域名解析有问题。
因为通常情况下我们会直接用git clone下载整个repository，很少下载单个文件，所以第一次遇到这个报错。需要以sudo权限打开/etc/hosts文件，sudo vim /etc/hosts，然后在末尾添加一行199.232.68.133 raw.githubusercontent.com。再次下载，成功。
看下脚本的用法。运行python脚本，生成我们自己DIY的Genomic Position File。默认使用gtf文件的gene_id字段，也可通过--attribute_name设置其它，例如gene_name字段。
python gtf_to_position_file.py -h
# By Default, gene_id
python gtf_to_position_file.py gencode.v32.primary_assembly.annotation.gtf gencode_v32_gene_pos_gene_id.txt
# gene_name
python gtf_to_position_file.py --attribute_name gene_name gencode.v32.primary_assembly.annotation.gtf gencode_v32_gene_pos_gene_name.txt
得到生成的文件如下，可以进行之后的infercnv步骤。
把源代码贴在下面学习一下，脚本相对比较简单，只调用了常规的argparse，csv，os三个标准库。写成一个函数convert_to_positional_file()解决问题。
#!/usr/bin/env python
"""
Converts GTF files to proprietary formats.
"""
# Import statements
import argparse
import csv
import os
__author__ = 'Timothy Tickle, Itay Tirosh, Brian Haas'
__copyright__ = 'Copyright 2016'
__credits__ = ["Timothy Tickle"]
__license__ = 'BSD-3'
__maintainer__ = 'Timothy Tickle'
__email__ = 'ttickle@bbroadinstitute.org'
__status__ = 'Development'
def convert_to_positional_file(input_gtf, output_positional, attribute_key):
""" Convert input GTF file to positional file.
:param input_gtf: Path to input gtf file
:type input_gtf: String
:param output_positional: Path to output positional file
:type output_positional: String
:param attribute_key: Key of the GTF attribute to use for feature/row names
:type attribute_key: String
:returns: Indicator of success (True) or Failure (False)
:rtype: boolean
"""
if not input_gtf or not os.path.exists(input_gtf):
print("".join(["gtf_to_position_file.py:: ",
"Could not find input file : " + input_gtf]))
all_genes_found = set()
# Holds lines to output after parsing.
output_line = []
previous_gene = None
previous_chr = None
gene_positions = []
# Metrics for the file
i_comments = 0
i_duplicate_entries = 0
i_entries = 0
i_accepted_entries = 0
i_written_lines = 0
with open(input_gtf, "r") as gtf:
gtf_file = csv.reader(gtf,delimiter="\t")
for gtf_line in gtf_file:
if gtf_line[0][0] == "#":
i_comments += 1
continue
i_entries += 1
# Clean up the attribute keys and match the one of interest.
attributes = gtf_line[8].split(";")
attributes = [entry.strip(" ") for entry in attributes]
attributes = [entry.split(" ") for entry in attributes if entry]
attributes = [[entry[0].strip('"'),entry[1].strip('"')] for entry in attributes]
attributes = dict([[entry[0].split("|")[0],entry[1]] for entry in attributes])
if attribute_key in attributes:
gene_name = attributes[attribute_key]
else:
print("Could not find an attribute in the GTF with the name '"+attribute_key+"'. Line="+"\t".join(gtf_line))
exit(99)
if not gene_name == previous_gene:
if len(gene_positions) > 1 and previous_gene not in all_genes_found:
i_accepted_entries += 1
gene_positions.sort()
output_line.append("\t".join([previous_gene,
previous_chr,
str(gene_positions[0]),
str(gene_positions[-1])]))
all_genes_found.add(previous_gene)
gene_positions = []
else:
i_duplicate_entries += 1
gene_positions += [int(gtf_line[3]), int(gtf_line[4])]
previous_gene = gene_name
previous_chr = gtf_line[0]
if previous_gene and previous_chr and len(gene_positions) > 1:
i_accepted_entries += 1
gene_positions.sort()
output_line.append("\t".join([previous_gene,
previous_chr,
str(gene_positions[0]),
str(gene_positions[-1])]))
with open(output_positional, "w") as positional_file:
i_written_lines += len(output_line)
positional_file.write("\n".join(output_line))
# Print metrics
print("Number of lines read: " + str(i_entries))
print("Number of comments: " + str(i_comments))
print("Number of entries: " + str(i_accepted_entries))
print("Number of duplicate entries: " + str(i_duplicate_entries))
print("Number of entries written: " + str(i_written_lines))
if __name__ == "__main__":
# Parse arguments
prsr_arguments = argparse.ArgumentParser(prog='gtf_to_position_file.py',
description='Convert a GTF file to a positional file.',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
# Add positional argument
prsr_arguments.add_argument("input_gtf",
metavar="input_gtf",
help="Path to the input GTF file.")
prsr_arguments.add_argument("--attribute_name",
metavar="attribute_name",
default="gene_id",
help="The name of the attribute in the GTF attributes to use instead of gene name, for example 'gene_name' or 'transcript_id'.")
prsr_arguments.add_argument("output_positional",
metavar="output_positional",
help="Path for the output positional file.")
args = prsr_arguments.parse_args()
# Run Script
convert_to_positional_file(args.input_gtf, args.output_positional, args.attribute_name)
