Updated vcf_filter

andrew/vcf_filter.py

@@ -49,16 +49,42 @@ def __call__(self, parser, args, value, option_string=None):
     # Other file arguments. Expand as needed
     vcf_parser.add_argument('--out', help = 'Specifies the output filename', type = str,  default = 'out', action = parser_confirm_no_file())
 
-    ### Site Filters
+    out_format_list = ['vcf', 'bcf']
+    out_format_default = 'bcf'
 
-    # Chromosome Filters
-    vcf_parser.add_argument('--filter-chr', help = 'Specifies the chromosome(s) to include', nargs = '+', type = str)
-    vcf_parser.add_argument('--filter-not-chr', help = 'Specifies the chromosome(s) to exclude', nargs = '+', type = str)
+    vcf_parser.add_argument('--out-format', metavar = '{' + ', '.join(out_format_list) + '}', help = 'Specifies the output format', type = str, choices = out_format_list, default = out_format_default)
+    ### Filters
 
-    # Position Filters
+    # Chromosome filters
+    vcf_parser.add_argument('--filter-include-chr', help = 'Specifies the chromosome(s) to include', nargs = '+', type = str)
+    vcf_parser.add_argument('--filter-exclude-chr', help = 'Specifies the chromosome(s) to exclude', nargs = '+', type = str)
+
+    # Basic position filters
     vcf_parser.add_argument('--filter-from-bp', help = 'Specifies the lower bound of sites to include (May only be used with a single chromosome)', type = int)
     vcf_parser.add_argument('--filter-to-bp', help = 'Specifies the upper bound of sites to include (May only be used with a single chromosome)', type = int)
 
+    # BED-based position filters
+    vcf_parser.add_argument('--filter-include-bed', help = 'Specifies a set of sites to include within a BED file', action = parser_confirm_file())
+    vcf_parser.add_argument('--filter-exclude-bed', help = 'Specifies a set of sites to exclude within a BED file', action = parser_confirm_file())
+
+    # Filter-flag filters
+    vcf_parser.add_argument('--filter-include-passed', help = "Specifies that only sites with the filter flag 'PASS' should be included", action = 'store_true')
+    vcf_parser.add_argument('--filter-include-filtered', help = 'Specifies that all sites with the given filter flag should be included', nargs = '+', type = str)
+    vcf_parser.add_argument('--filter-exclude-filtered', help = 'Specifies that all sites with the given filter flag should be excluded', nargs = '+', type = str)
+
+    # Info-flag filters
+    vcf_parser.add_argument('--filter-include-info', help = 'Specifies that all sites with the given info flag should be included', nargs = '+', type = str)
+    vcf_parser.add_argument('--filter-exclude-info', help = 'Specifies that all sites with the given info flag should be excluded', nargs = '+', type = str)
+
+    # Allele count filters
+    vcf_parser.add_argument('--filter-min-alleles', help = 'Specifies that only sites with a number of allele >= to the number given should be included', type = int)
+    vcf_parser.add_argument('--filter-max-alleles', help = 'Specifies that only sites with a number of allele <= to the number given should be included', type = int)
+
+    # Missing data filter
+    vcf_parser.add_argument('--filter-max-missing', help = 'Specifies that only sites with more than this number of genotypes among individuals should be included', type = int)
+
+    # Additional Filters
+    vcf_parser.add_argument('--filter-distance', help = 'Specifies a distance that no two sites may be within', type = int)
 
     if passed_arguments:
         return vcf_parser.parse_args(passed_arguments)
@@ -98,69 +124,64 @@ def run (passed_arguments = []):
     # Grab VCF arguments from command line
     vcf_args = vcf_filter_parser(passed_arguments)
 
-    print vcf_args
-
-    '''
     # Argument container for vcftools
     vcftools_call_args = ['--out', vcf_args.out]
 
-    if vcf_args.calc_statistic == 'windowed-weir-fst':
-        # Confirms that at least two population files have been specified
-        if not vcf_args.pop_file or len(vcf_args.pop_file) < 2:
-            sys.exit('Two or more population files requried. Please assign using --pop-file')
-
-        # Assigns specific vcftools arguments for calculating fst
-        vcftools_pop_args = [population_args for population_file in vcf_args.pop_file for population_args in ['--weir-fst-pop', population_file]]
-        vcftools_window_args = ['--fst-window-size', vcf_args.statistic_window_size, '--fst-window-step', vcf_args.statistic_window_step]
-
-        # Assigns all the vcftools arguments for calculating windowed fst
-        vcftools_call_args.extend(vcftools_pop_args + vcftools_window_args)
-
-        # Assigns the suffix for the vcftools log file
-        vcftools_log_suffix = '.windowed.weir.fst'
-
-    elif vcf_args.calc_statistic == 'weir-fst':
-        # Confirms that at least two population files have been specified
-        if not vcf_args.pop_file or len(vcf_args.pop_file) < 2:
-            sys.exit('Two or more population files requried. Please assign using --pop-file')
-
-        # Assigns specific vcftools arguments for calculating site-based fst
-        vcftools_call_args.extend([population_args for population_file in vcf_args.pop_file for population_args in ['--weir-fst-pop', population_file]])
-
-        # Assigns the suffix for the vcftools log file
-        vcftools_log_suffix = '.weir.fst'
-
-    elif vcf_args.calc_statistic == 'TajimaD':
-
-        # Assigns all the vcftools arguments for calculating TajimaD
-        vcftools_call_args.extend(['--TajimaD', vcf_args.statistic_window_size])
-
-        # Assigns the suffix for the vcftools log file
-        vcftools_log_suffix = '.Tajima.D'
-
-    elif vcf_args.calc_statistic == 'pi':
-
-        # Assigns all the vcftools arguments for calculating pi
-        vcftools_call_args.extend(['--window-pi', vcf_args.statistic_window_size, '--window-pi-step', vcf_args.statistic_window_step])
-
-        # Assigns the suffix for the vcftools log file
-        vcftools_log_suffix = '.windowed.pi'
-
-    elif vcf_args.calc_statistic == 'freq':
-
-        # Assigns all the vcftools arguments for the allele frequency
-        vcftools_call_args.extend(['--freq'])
-
-        # Assigns the suffix for the vcftools log file
-        vcftools_log_suffix = '.frq'
-
-    elif vcf_args.calc_statistic == 'het':
-
-        # Assigns all the vcftools arguments for calculating heterozygosity
-        vcftools_call_args.extend(['--het'])
-
-        # Assigns the suffix for the vcftools log file
-        vcftools_log_suffix = '.het'
+    if vcf_args.out_format:
+        if vcf_args.out_format == 'bcf':
+            vcftools_call_args.append('--recode-bcf')
+        elif vcf_args.out_format == 'vcf':
+            vcftools_call_args.append('--recode')
+
+    if vcf_args.filter_include_chr or vcf_args.filter_exclude_chr:
+        if vcf_args.filter_include_chr:
+            for chr_to_include in vcf_args.filter_include_chr:
+                vcftools_call_args.extend(['--chr', chr_to_include])
+        if vcf_args.filter_exclude_chr:
+            for chr_to_exclude in vcf_args.filter_exclude_chr:
+                vcftools_call_args.extend(['--not-chr', chr_to_exclude])
+
+    if vcf_args.filter_from_bp or vcf_args.filter_to_bp:
+        if vcf_args.filter_include_chr:
+            vcftools_call_args.extend(['--from-bp', vcf_args.filter_from_bp])
+        if vcf_args.filter_exclude_chr:
+            vcftools_call_args.extend(['--to-bp', vcf_args.filter_to_bp])
+
+    if vcf_args.filter_include_bed or vcf_args.filter_exclude_bed:
+        if vcf_args.filter_include_bed:
+            vcftools_call_args.extend(['--bed', vcf_args.filter_include_bed])
+        if vcf_args.filter_exclude_bed:
+            vcftools_call_args.extend(['--exclude-bed', vcf_args.filter_exclude_bed])
+
+    if vcf_args.filter_include_passed or vcf_args.filter_include_filtered or vcf_args.filter_exclude_filtered:
+        if vcf_args.filter_include_passed:
+            vcftools_call_args.append('--remove-filtered-all')
+        if vcf_args.filter_include_filtered:
+            for filtered_to_include in vcf_args.filter_include_filtered:
+                vcftools_call_args.extend(['--keep-filtered', filtered_to_include])
+        if vcf_args.filter_exclude_filtered:
+            for filtered_to_exclude in vcf_args.filter_exclude_filtered:
+                vcftools_call_args.extend(['--remove-filtered', filtered_to_exclude])
+
+    if vcf_args.filter_include_info or vcf_args.filter_exclude_info:
+        if vcf_args.filter_include_info:
+            for info_to_include in vcf_args.filter_include_info:
+                vcftools_call_args.extend(['--keep-INFO', info_to_include])
+        if vcf_args.filter_exclude_info:
+            for info_to_exclude in vcf_args.filter_exclude_info:
+                vcftools_call_args.extend(['--remove-INFO', info_to_exclude])
+
+    if vcf_args.filter_min_alleles or vcf_args.filter_max_alleles:
+        if vcf_args.filter_min_alleles:
+            vcftools_call_args.extend(['--min-alleles', vcf_args.filter_min_alleles])
+        if vcf_args.filter_max_alleles:
+            vcftools_call_args.extend(['--max-alleles', vcf_args.filter_max_alleles])
+
+    if vcf_args.filter_max_missing:
+        vcftools_call_args.extend(['--max-missing-count', vcf_args.filter_max_missing])
+
+    if vcf_args.filter_distance:
+        vcftools_call_args.extend(['--thin', vcf_args.filter_distance])
 
     # Assigns the file argument for vcftools
     vcfname_arg = assign_vcftools_input_arg(vcf_args.vcfname)
@@ -171,8 +192,8 @@ def run (passed_arguments = []):
 
     # Check that the log file was created correctly, get the suffix for the log file, and create the file
     if check_vcftools_for_errors(vcftools_err):
-        produce_vcftools_log(vcftools_err, vcf_args.out, vcftools_log_suffix)
-    '''
+        produce_vcftools_log(vcftools_err, vcf_args.out, '.filter')
+
 
 if __name__ == "__main__":
     #initLogger()