From 9b4f6afa212f838504f002469e080ebe274ab59f Mon Sep 17 00:00:00 2001 From: Christopher Hartl Date: Fri, 20 Jan 2012 23:07:59 -0500 Subject: [PATCH] Alterations to scripts for better performance. Grid search now expands the sens/spec tradeoff (90 was far too aggressive against hapmap chr20), and 20 max gaussians was too many, and caused errors. For consensus genotypes: remember to gunzip the beagle outputs before converting to VCF. Also, beagle can in fact create 'null' alleles in certain circumstances. I'm not sure what exactly those circumstances are, but those sites should be ignored. When it does, all alleles apear to be set to null, so this should not affect the actual phasing in the output VCF. --- .../sting/gatk/walkers/beagle/BeagleOutputToVCFWalker.java | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/public/java/src/org/broadinstitute/sting/gatk/walkers/beagle/BeagleOutputToVCFWalker.java b/public/java/src/org/broadinstitute/sting/gatk/walkers/beagle/BeagleOutputToVCFWalker.java index f827856be..4f115b46e 100755 --- a/public/java/src/org/broadinstitute/sting/gatk/walkers/beagle/BeagleOutputToVCFWalker.java +++ b/public/java/src/org/broadinstitute/sting/gatk/walkers/beagle/BeagleOutputToVCFWalker.java @@ -241,6 +241,11 @@ public class BeagleOutputToVCFWalker extends RodWalker { String alleleA = beagleGenotypePairs.get(0); String alleleB = beagleGenotypePairs.get(1); + if ( alleleA.equals("null") || alleleB.equals("null") ) { + logger.warn("Beagle produced 'null' alleles at location "+ref.getLocus().toString()+". Ignoring."); + return 0; + } + // Beagle always produces genotype strings based on the strings we input in the likelihood file. String refString = vc_input.getReference().getDisplayString(); if (refString.length() == 0) // ref was null