Merge pull request #82 from hall-lab/max_ci_dist

Stub in code for toggling use of 95 pct confidence interval

svtyper/classic.py

@@ -9,7 +9,7 @@
 
 import svtyper.version
 
-from svtyper.parsers import Vcf, Variant, Sample
+from svtyper.parsers import Vcf, Variant, Sample, confidence_interval
 from svtyper.utils import *
 from svtyper.statistics import bayes_gt
 
@@ -32,6 +32,7 @@ def get_args():
     parser.add_argument('-n', dest='num_samp', metavar='INT', type=int, required=False, default=1000000, help='number of reads to sample from BAM file for building insert size distribution [1000000]')
     parser.add_argument('-q', '--sum_quals', action='store_true', required=False, help='add genotyping quality to existing QUAL (default: overwrite QUAL field)')
     parser.add_argument('--max_reads', metavar='INT', type=int, default=None, required=False, help='maximum number of reads to assess at any variant (reduces processing time in high-depth regions, default: unlimited)')
+    parser.add_argument('--max_ci_dist', metavar='INT', type=int, default=1e10, required=False, help='maximum size of a confidence interval before 95% CI is used intead (default: 1e10)')
     parser.add_argument('--split_weight', metavar='FLOAT', type=float, required=False, default=1, help='weight for split reads [1]')
     parser.add_argument('--disc_weight', metavar='FLOAT', type=float, required=False, default=1, help='weight for discordant paired-end reads [1]')
     parser.add_argument('-w', '--write_alignment', metavar='FILE', dest='alignment_outpath', type=str, required=False, default=None, help='write relevant reads to BAM file')
@@ -115,7 +116,8 @@ def sv_genotype(bam_string,
                 alignment_outpath,
                 ref_fasta,
                 sum_quals,
-                max_reads):
+                max_reads,
+                max_ci_dist):
 
     # parse the comma separated inputs
     bam_list = []
@@ -236,8 +238,8 @@ def sv_genotype(bam_string,
                 posA = var.pos
                 posB = var2.pos
                 # confidence intervals
-                ciA = map(int, var.info['CIPOS'].split(','))
-                ciB = map(int, var2.info['CIPOS'].split(','))
+                ciA = confidence_interval(var, 'CIPOS', 'CIPOS95', max_ci_dist)
+                ciB = confidence_interval(var2, 'CIPOS', 'CIPOS95', max_ci_dist)
 
                 # infer the strands from the alt allele
                 if var.alt[-1] == '[' or var.alt[-1] == ']':
@@ -259,8 +261,8 @@ def sv_genotype(bam_string,
             posA = var.pos
             posB = int(var.get_info('END'))
             # confidence intervals
-            ciA = map(int, var.info['CIPOS'].split(','))
-            ciB = map(int, var.info['CIEND'].split(','))
+            ciA = confidence_interval(var, 'CIPOS', 'CIPOS95', max_ci_dist)
+            ciB = confidence_interval(var, 'CIEND', 'CIEND95', max_ci_dist)
             if svtype == 'DEL':
                 var_length = posB - posA
                 o1_is_reverse, o2_is_reverse =  False, True
@@ -561,7 +563,8 @@ def main():
                 args.alignment_outpath,
                 args.ref_fasta,
                 args.sum_quals,
-                args.max_reads)
+                args.max_reads,
+                args.max_ci_dist)
 
 # --------------------------------------
 # command-line/console entrypoint

svtyper/parsers.py

@@ -8,6 +8,12 @@
 # ==================================================
 # VCF parsing tools
 # ==================================================
+def confidence_interval(var, tag, alt_tag, max_ci_dist):
+    ci = map(int, var.info[tag].split(','))
+    if ci[1] - ci[0] > max_ci_dist:
+        return map(int, var.info[alt_tag].split(','))
+    return ci
+
 
 class Vcf(object):
     def __init__(self):
@@ -120,7 +126,7 @@ def sample_to_col(self, sample):
     def _init_bnd_breakpoint_func():
         bnd_cache = {}
 
-        def _get_bnd_breakpoints(variant):
+        def _get_bnd_breakpoints(variant, max_ci_dist):
             if variant.info['MATEID'] in bnd_cache:
                 var2 = variant
                 var = bnd_cache[variant.info['MATEID']]
@@ -129,8 +135,8 @@ def _get_bnd_breakpoints(variant):
                 posA = var.pos
                 posB = var2.pos
                 # confidence intervals
-                ciA = map(int, var.info['CIPOS'].split(','))
-                ciB = map(int, var2.info['CIPOS'].split(','))
+                ciA = confidence_interval(var, 'CIPOS', 'CIPOS95', max_ci_dist)
+                ciB = confidence_interval(var2, 'CIPOS', 'CIPOS95', max_ci_dist)
 
                 # infer the strands from the alt allele
                 if var.alt[-1] == '[' or var.alt[-1] == ']':
@@ -163,14 +169,14 @@ def _get_bnd_breakpoints(variant):
         return _get_bnd_breakpoints
 
     @staticmethod
-    def _default_get_breakpoints(variant):
+    def _default_get_breakpoints(variant, max_ci_dist):
         chromA = variant.chrom
         chromB = variant.chrom
         posA = variant.pos
         posB = int(variant.get_info('END'))
         # confidence intervals
-        ciA = map(int, variant.info['CIPOS'].split(','))
-        ciB = map(int, variant.info['CIEND'].split(','))
+        ciA = confidence_interval(variant, 'CIPOS', 'CIPOS95', max_ci_dist)
+        ciB = confidence_interval(variant, 'CIEND', 'CIEND95', max_ci_dist)
         svtype = variant.get_svtype()
         if svtype == 'DEL':
             var_length = posB - posA
@@ -202,17 +208,17 @@ def _default_get_breakpoints(variant):
 
         return breakpoints
 
-    def get_variant_breakpoints(self, variant):
+    def get_variant_breakpoints(self, variant, max_ci_dist):
         if self._bnd_breakpoint_func is None:
             func = self._init_bnd_breakpoint_func()
             self._bnd_breakpoint_func = func
 
         breakpoints = None
         if variant.get_svtype() == 'BND':
             func = self._bnd_breakpoint_func
-            breakpoints = func(variant)
+            breakpoints = func(variant, max_ci_dist)
         else:
-            breakpoints = self._default_get_breakpoints(variant)
+            breakpoints = self._default_get_breakpoints(variant, max_ci_dist)
 
         return breakpoints
 

svtyper/singlesample.py

@@ -5,7 +5,7 @@
 from cytoolz.itertoolz import partition_all
 
 import svtyper.version
-from svtyper.parsers import Vcf, Variant, Sample, SamFragment
+from svtyper.parsers import Vcf, Variant, Sample, SamFragment, confidence_interval
 from svtyper.utils import die, logit, prob_mapq, write_sample_json, tempdir, vcf_headers, vcf_variants, vcf_samples
 from svtyper.statistics import bayes_gt
 
@@ -28,6 +28,7 @@ def get_args():
     parser.add_argument('-n', dest='num_samp', metavar='INT', type=int, required=False, default=1000000, help='number of reads to sample from BAM file for building insert size distribution [1000000]')
     parser.add_argument('-q', '--sum_quals', action='store_true', required=False, help='add genotyping quality to existing QUAL (default: overwrite QUAL field)')
     parser.add_argument('--max_reads', metavar='INT', type=int, default=1000, required=False, help='maximum number of reads to assess at any variant (reduces processing time in high-depth regions, default: 1000)')
+    parser.add_argument('--max_ci_dist', metavar='INT', type=int, default=1e10, required=False, help='maximum size of a confidence interval before 95% CI is used intead (default: 1e10)')
     parser.add_argument('--split_weight', metavar='FLOAT', type=float, required=False, default=1, help='weight for split reads [1]')
     parser.add_argument('--disc_weight', metavar='FLOAT', type=float, required=False, default=1, help='weight for discordant paired-end reads [1]')
     parser.add_argument('--debug', action='store_true', help=argparse.SUPPRESS)
@@ -123,14 +124,14 @@ def init_vcf(vcffile, sample, scratchdir):
     v.add_sample(sample.name)
     return v
 
-def collect_breakpoints(vcf):
+def collect_breakpoints(vcf, max_ci_dist):
     breakpoints = []
     for vline in vcf_variants(vcf.filename):
         v = vline.rstrip().split('\t')
         variant = Variant(v, vcf)
         if not variant.has_svtype(): continue
         if not variant.is_valid_svtype(): continue
-        brkpts = vcf.get_variant_breakpoints(variant)
+        brkpts = vcf.get_variant_breakpoints(variant, max_ci_dist)
         if brkpts is None: continue
         breakpoints.append(brkpts)
     return breakpoints
@@ -573,7 +574,7 @@ def assign_genotype_to_variant(variant, sample, genotype_result):
         variant.genotype(sample.name).set_format('AB',  outcome['formats']['AB'])
     return variant
 
-def genotype_serial(src_vcf, out_vcf, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, debug):
+def genotype_serial(src_vcf, out_vcf, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, max_ci_dist, debug):
     # initializations
     bnd_cache = {}
     src_vcf.write_header(out_vcf)
@@ -607,7 +608,7 @@ def genotype_serial(src_vcf, out_vcf, sample, z, split_slop, min_aligned, sum_qu
             variant.write(out_vcf)
             continue
 
-        breakpoints = src_vcf.get_variant_breakpoints(variant)
+        breakpoints = src_vcf.get_variant_breakpoints(variant, max_ci_dist)
 
         # special BND processing
         if variant.get_svtype() == 'BND':
@@ -706,15 +707,15 @@ def apply_genotypes_to_vcf(src_vcf, out_vcf, genotypes, sample, sum_quals):
             variant2.genotype = variant.genotype
             variant2.write(out_vcf)
 
-def genotype_parallel(src_vcf, out_vcf, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, debug, cores, breakpoint_batch_size, ref_fasta):
+def genotype_parallel(src_vcf, out_vcf, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, max_ci_dist, debug, cores, breakpoint_batch_size, ref_fasta):
 
     # cleanup unused library attributes
     for rg in sample.rg_to_lib:
         sample.rg_to_lib[rg].cleanup()
 
     # 1st pass through input vcf -- collect all the relevant breakpoints
     logit("Collecting breakpoints")
-    breakpoints = collect_breakpoints(src_vcf)
+    breakpoints = collect_breakpoints(src_vcf, max_ci_dist)
     logit("Number of breakpoints/SVs to process: {}".format(len(breakpoints)))
     logit("Collecting regions")
     regions = [ get_breakpoint_regions(b, sample, z) for b in breakpoints ]
@@ -772,6 +773,7 @@ def sso_genotype(bam_string,
                  ref_fasta,
                  sum_quals,
                  max_reads,
+                 max_ci_dist,
                  cores,
                  batch_size):
 
@@ -802,11 +804,11 @@ def sso_genotype(bam_string,
         if cores is None:
             logit("Genotyping Input VCF (Serial Mode)")
             # pass through input vcf -- perform actual genotyping
-            genotype_serial(src_vcf, vcf_out, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, debug)
+            genotype_serial(src_vcf, vcf_out, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, max_ci_dist, debug)
         else:
             logit("Genotyping Input VCF (Parallel Mode)")
 
-            genotype_parallel(src_vcf, vcf_out, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, debug, cores, batch_size, ref_fasta)
+            genotype_parallel(src_vcf, vcf_out, sample, z, split_slop, min_aligned, sum_quals, split_weight, disc_weight, max_reads, max_ci_dist, debug, cores, batch_size, ref_fasta)
 
 
     sample.close()
@@ -834,6 +836,7 @@ def main():
                  args.ref_fasta,
                  args.sum_quals,
                  args.max_reads,
+                 args.max_ci_dist,
                  args.cores,
                  args.batch_size)
 

svtyper/version.py

@@ -1,2 +1,2 @@
 __author__ = "Colby Chiang (colbychiang@wustl.edu)"
-__version__ = "v0.6.0"
+__version__ = "v0.6.2"

tests/test_singlesample.py

@@ -31,6 +31,7 @@ def test_serial_integration(self):
                            ref_fasta=None,
                            sum_quals=False,
                            max_reads=1000,
+                           max_ci_dist=1e10,
                            cores=None,
                            batch_size=1000)
 
@@ -56,6 +57,7 @@ def test_parallel_integration(self):
                            ref_fasta=None,
                            sum_quals=False,
                            max_reads=1000,
+                           max_ci_dist=1e10,
                            cores=1,
                            batch_size=1000)
 

tests/test_svtyper.py

@@ -77,7 +77,8 @@ def test_integration(self):
                              alignment_outpath=None,
                              ref_fasta=None,
                              sum_quals=False,
-                             max_reads=None)
+                             max_reads=None,
+                             max_ci_dist=1e10)
 
         fail_msg = "did not file output vcf '{}' after running sv_genotype".format(out_vcf)
         self.assertTrue(os.path.exists(out_vcf), fail_msg)