Matched gCNV pipeline arguments to those that were shown to have good…

… performance in running large exome cohorts. (#8234)

scripts/cnv_wdl/cnv_common_tasks.wdl

@@ -158,8 +158,8 @@ task FilterIntervals {
             --maximum-mappability ~{default="1.0" maximum_mappability} \
             --minimum-segmental-duplication-content ~{default="0.0" minimum_segmental_duplication_content} \
             --maximum-segmental-duplication-content ~{default="0.5" maximum_segmental_duplication_content} \
-            --low-count-filter-count-threshold ~{default="5" low_count_filter_count_threshold} \
-            --low-count-filter-percentage-of-samples ~{default="90.0" low_count_filter_percentage_of_samples} \
+            --low-count-filter-count-threshold ~{default="10" low_count_filter_count_threshold} \
+            --low-count-filter-percentage-of-samples ~{default="50.0" low_count_filter_percentage_of_samples} \
             --extreme-count-filter-minimum-percentile ~{default="1.0" extreme_count_filter_minimum_percentile} \
             --extreme-count-filter-maximum-percentile ~{default="99.0" extreme_count_filter_maximum_percentile} \
             --extreme-count-filter-percentage-of-samples ~{default="90.0" extreme_count_filter_percentage_of_samples} \

scripts/cnv_wdl/germline/cnv_germline_case_workflow.wdl

@@ -323,7 +323,7 @@ task DetermineGermlineContigPloidyCaseMode {
             --output ~{output_dir_} \
             --output-prefix case \
             --verbosity DEBUG \
-            --mapping-error-rate ~{default="0.01" mapping_error_rate} \
+            --mapping-error-rate ~{default="0.3" mapping_error_rate} \
             --sample-psi-scale ~{default="0.0001" sample_psi_scale}
 
         tar c -C ~{output_dir_}/case-calls . | gzip -1 > case-contig-ploidy-calls.tar.gz
@@ -423,11 +423,11 @@ task GermlineCNVCallerCaseMode {
             --output ~{output_dir_} \
             --output-prefix case \
             --verbosity DEBUG \
-            --p-alt ~{default="1e-6" p_alt} \
+            --p-alt ~{default="5e-4" p_alt} \
             --cnv-coherence-length ~{default="10000.0" cnv_coherence_length} \
             --max-copy-number ~{default="5" max_copy_number} \
             --mapping-error-rate ~{default="0.01" mapping_error_rate} \
-            --sample-psi-scale ~{default="0.0001" sample_psi_scale} \
+            --sample-psi-scale ~{default="0.01" sample_psi_scale} \
             --depth-correction-tau ~{default="10000.0" depth_correction_tau} \
             --copy-number-posterior-expectation-mode ~{default="HYBRID" copy_number_posterior_expectation_mode} \
             --active-class-padding-hybrid-mode ~{default="50000" active_class_padding_hybrid_mode} \

scripts/cnv_wdl/germline/cnv_germline_cohort_workflow.wdl

@@ -485,8 +485,8 @@ task DetermineGermlineContigPloidyCohortMode {
             --output ~{output_dir_} \
             --output-prefix ~{cohort_entity_id} \
             --verbosity DEBUG \
-            --mean-bias-standard-deviation ~{default="0.01" mean_bias_standard_deviation} \
-            --mapping-error-rate ~{default="0.01" mapping_error_rate} \
+            --mean-bias-standard-deviation ~{default="1" mean_bias_standard_deviation} \
+            --mapping-error-rate ~{default="0.3" mapping_error_rate} \
             --global-psi-scale ~{default="0.001" global_psi_scale} \
             --sample-psi-scale ~{default="0.0001" sample_psi_scale}
 
@@ -599,15 +599,15 @@ task GermlineCNVCallerCohortMode {
             --output ~{output_dir_} \
             --output-prefix ~{cohort_entity_id} \
             --verbosity DEBUG \
-            --p-alt ~{default="1e-6" p_alt} \
-            --p-active ~{default="1e-2" p_active} \
+            --p-alt ~{default="5e-4" p_alt} \
+            --p-active ~{default="1e-1" p_active} \
             --cnv-coherence-length ~{default="10000.0" cnv_coherence_length} \
             --class-coherence-length ~{default="10000.0" class_coherence_length} \
             --max-copy-number ~{default="5" max_copy_number} \
-            --max-bias-factors ~{default="5" max_bias_factors} \
+            --max-bias-factors ~{default="6" max_bias_factors} \
             --mapping-error-rate ~{default="0.01" mapping_error_rate} \
-            --interval-psi-scale ~{default="0.001" interval_psi_scale} \
-            --sample-psi-scale ~{default="0.0001" sample_psi_scale} \
+            --interval-psi-scale ~{default="0.01" interval_psi_scale} \
+            --sample-psi-scale ~{default="0.01" sample_psi_scale} \
             --depth-correction-tau ~{default="10000.0" depth_correction_tau} \
             --log-mean-bias-standard-deviation ~{default="0.1" log_mean_bias_standard_deviation} \
             --init-ard-rel-unexplained-variance ~{default="0.1" init_ard_rel_unexplained_variance} \

src/main/java/org/broadinstitute/hellbender/tools/copynumber/FilterIntervals.java

@@ -229,7 +229,7 @@ public final class FilterIntervals extends CommandLineProgram {
             minValue = 0,
             optional = true
     )
-    private int lowCountFilterCountThreshold = 5;
+    private int lowCountFilterCountThreshold = 10;
 
     @Argument(
             doc = "Percentage-of-samples parameter for the low-count filter.  Intervals with a count " +
@@ -241,7 +241,7 @@ public final class FilterIntervals extends CommandLineProgram {
             maxValue = 100.,
             optional = true
     )
-    private double lowCountFilterPercentageOfSamples = 90.;
+    private double lowCountFilterPercentageOfSamples = 50.;
 
     @Argument(
             doc = "Minimum-percentile parameter for the extreme-count filter.  Intervals with a count " +

src/main/java/org/broadinstitute/hellbender/tools/copynumber/arguments/GermlineCallingArgumentCollection.java

@@ -37,7 +37,7 @@ public final class GermlineCallingArgumentCollection implements Serializable {
             maxValue = 1.,
             optional = true
     )
-    private double pAlt = 1e-6;
+    private double pAlt = 5e-4;
 
     @Argument(
             doc = "Prior probability of treating an interval as CNV-active (in a CNV-active domains, all " +
@@ -47,7 +47,7 @@ public final class GermlineCallingArgumentCollection implements Serializable {
             maxValue = 1.,
             optional = true
     )
-    private double pActive = 1e-2;
+    private double pActive = 1e-1;
 
     @Argument(
             doc = "Coherence length of CNV events (in the units of bp).",

src/main/java/org/broadinstitute/hellbender/tools/copynumber/arguments/GermlineContigPloidyModelArgumentCollection.java

@@ -27,7 +27,7 @@ public final class GermlineContigPloidyModelArgumentCollection implements Serial
             minValue = 0.,
             optional = true
     )
-    private double meanBiasStandardDeviation = 0.01;
+    private double meanBiasStandardDeviation = 1.;
 
     @Argument(
             doc = "Typical mapping error rate.",
@@ -36,7 +36,7 @@ public final class GermlineContigPloidyModelArgumentCollection implements Serial
             maxValue = 1.,
             optional = true
     )
-    private double mappingErrorRate = 0.01;
+    private double mappingErrorRate = 0.3;
 
     @Argument(
             doc = "Prior scale of contig coverage unexplained variance.  If a single sample is provided, " +

src/main/java/org/broadinstitute/hellbender/tools/copynumber/arguments/GermlineDenoisingModelArgumentCollection.java

@@ -61,7 +61,7 @@ public enum CopyNumberPosteriorExpectationMode {
             minValue = 0,
             optional = true
     )
-    private int maxBiasFactors = 5;
+    private int maxBiasFactors = 6;
 
     @Argument(
             doc = "Typical mapping error rate.",
@@ -78,15 +78,15 @@ public enum CopyNumberPosteriorExpectationMode {
             minValue = 0.,
             optional = true
     )
-    private double intervalPsiScale = 0.001;
+    private double intervalPsiScale = 0.01;
 
     @Argument(
             doc = "Typical scale of sample-specific correction to the unexplained variance.",
             fullName = SAMPLE_PSI_SCALE_LONG_NAME,
             minValue = 0.,
             optional = true
     )
-    private double samplePsiScale = 0.0001;
+    private double samplePsiScale = 0.01;
 
     @Argument(
             doc = "Precision of read depth pinning to its global value.",

src/main/python/org/broadinstitute/hellbender/gcnvkernel/models/model_denoising_calling.py

@@ -36,10 +36,10 @@ class DenoisingModelConfig:
     _q_c_expectation_modes = ['map', 'exact', 'hybrid']
 
     def __init__(self,
-                 max_bias_factors: int = 5,
+                 max_bias_factors: int = 6,
                  mapping_error_rate: float = 0.01,
-                 psi_t_scale: float = 0.001,
-                 psi_s_scale: float = 0.0001,
+                 psi_t_scale: float = 0.01,
+                 psi_s_scale: float = 0.01,
                  depth_correction_tau: float = 10000.0,
                  log_mean_bias_std: float = 0.1,
                  init_ard_rel_unexplained_variance: float = 0.1,
@@ -202,8 +202,8 @@ def from_json_file(json_file: str) -> 'DenoisingModelConfig':
 class CopyNumberCallingConfig:
     """Configuration of the copy number caller."""
     def __init__(self,
-                 p_alt: float = 1e-6,
-                 p_active: float = 1e-3,
+                 p_alt: float = 5e-4,
+                 p_active: float = 0.1,
                  cnv_coherence_length: float = 10000.0,
                  class_coherence_length: float = 10000.0,
                  max_copy_number: int = 5,

src/main/python/org/broadinstitute/hellbender/gcnvkernel/models/model_ploidy.py

@@ -24,10 +24,10 @@ class PloidyModelConfig:
     """Germline contig ploidy model hyper-parameters."""
     def __init__(self,
                  contig_ploidy_prior_map: Dict[str, np.ndarray] = None,
-                 mean_bias_sd: float = 1e-2,
+                 mean_bias_sd: float = 1.,
                  psi_j_scale: float = 1e-3,
                  psi_s_scale: float = 1e-4,
-                 mapping_error_rate: float = 1e-2):
+                 mapping_error_rate: float = 0.3):
         """Initializer.
 
         Args: