Mercurial > repos > miller-lab > genome_diversity
view add_fst_column.py @ 33:5064f618ec1c
remove munkres dependency
author | Richard Burhans <burhans@bx.psu.edu> |
---|---|
date | Fri, 20 Sep 2013 14:01:30 -0400 |
parents | 8997f2ca8c7a |
children |
line wrap: on
line source
#!/usr/bin/env python import gd_util import sys from Population import Population ################################################################################ if len(sys.argv) != 13: gd_util.die('Usage') input, p1_input, p2_input, input_type, data_source, min_reads, min_qual, retain, discard_fixed, biased, output, ind_arg = sys.argv[1:] p_total = Population() p_total.from_wrapped_dict(ind_arg) p1 = Population() p1.from_population_file(p1_input) if not p_total.is_superset(p1): gd_util.die('There is an individual in population 1 that is not in the SNP table') p2 = Population() p2.from_population_file(p2_input) if not p_total.is_superset(p2): gd_util.die('There is an individual in population 2 that is not in the SNP table') ################################################################################ prog = 'Fst_column' args = [ prog ] args.append(input) args.append(data_source) args.append(min_reads) args.append(min_qual) args.append(retain) args.append(discard_fixed) args.append(biased) columns = p1.column_list() for column in columns: if input_type == 'gd_genotype': column = int(column) - 2 args.append('{0}:1'.format(column)) columns = p2.column_list() for column in columns: if input_type == 'gd_genotype': column = int(column) - 2 args.append('{0}:2'.format(column)) with open(output, 'w') as fh: gd_util.run_program(prog, args, stdout=fh) sys.exit(0)