Restructured AlleleFrequencyMetricsWalker to correctly report Hapmap concordance numbers for genotyping and added reporting for Hapmap reference/variant calling. Also, tiny bugfix in interval code.

git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@181 348d0f76-0448-11de-a6fe-93d51630548a

java/src/org/broadinstitute/sting/gatk/TraversalEngine.java

@@ -597,7 +597,7 @@ public class TraversalEngine {
 
             // if we don't have a particular interval we're processing, check them all, otherwise only operate at this
             // location
-            if ( ( interval == null && inLocations(locus.getLocation()) ) || interval.overlapsP(locus.getLocation()) )  {
+            if ( ( interval == null && inLocations(locus.getLocation()) ) || (interval != null && interval.overlapsP(locus.getLocation())) )  {
 
                 //System.out.format("Working at %s\n", locus.getLocation().toString());
 

java/src/org/broadinstitute/sting/playground/gatk/walkers/AlleleFrequencyMetricsWalker.java

@@ -26,8 +26,10 @@ public class AlleleFrequencyMetricsWalker extends BasicLociWalker<AlleleFrequenc
     long num_loci_total=0;
     long num_loci_confident=0;
     double LOD_cutoff = 5;
-    long hapmap_tp=0;
+    long hapmap_genotype_correct = 0;
-    long hapmap_fp=0;
+    long hapmap_genotype_incorrect = 0;
+    long hapmap_refvar_correct = 0;
+    long hapmap_refvar_incorrect = 0;
 
     AlleleFrequencyWalker caller;
 
@@ -37,16 +39,9 @@ public class AlleleFrequencyMetricsWalker extends BasicLociWalker<AlleleFrequenc
 
         num_loci_total += 1;
 
-        if (Math.abs(alleleFreq.LOD) >= LOD_cutoff) { num_loci_confident += 1; }
-
-        if (alleleFreq.getQstar() > 0.0 && alleleFreq.getLOD() >= LOD_cutoff)
-        { 
-            // Confident variant.
-           
-            num_variants += 1;
-
         boolean is_dbSNP_SNP = false;
-            boolean hapmap_hit = false;
+        boolean has_hapmap_chip_genotype = false;
+        rodGFF hapmap_chip_genotype = null;
 
         for ( ReferenceOrderedDatum datum : rodData )
         {
@@ -60,20 +55,19 @@ public class AlleleFrequencyMetricsWalker extends BasicLociWalker<AlleleFrequenc
 
                 if ( datum instanceof rodGFF )
                 {
-                        rodGFF hapmap = (rodGFF) datum;
+                    has_hapmap_chip_genotype = true;
-                        String hapmap_genotype = hapmap.getFeature();
+                    hapmap_chip_genotype = (rodGFF)datum;
-                        String called_genotype = alleleFreq.asString();
+                }
-                        System.out.format("HAPMAP %s %s %.2f\n", hapmap_genotype, called_genotype, alleleFreq.getLOD());
+            }
-                        // There is a hapmap site here, is it correct?
+        }
-                        if (hapmap_genotype.equals(called_genotype))
+
+        if (Math.abs(alleleFreq.LOD) >= LOD_cutoff) { num_loci_confident += 1; }
+
+        if (alleleFreq.qstar > 0.0 && alleleFreq.LOD >= LOD_cutoff)
         { 
-                            hapmap_tp++;
+            // Confident variant.
-                        } else {
+           
-                            hapmap_fp++;
+            num_variants += 1;
-                        }
-                    }
-                }
-            }
 
             if (is_dbSNP_SNP)
             {
@@ -81,6 +75,75 @@ public class AlleleFrequencyMetricsWalker extends BasicLociWalker<AlleleFrequenc
             }
         }
 
+        // Convert qstar and LOD to give best qstar and a positive LOD for that qstar
+        double qstar;
+        double LOD;
+        if (alleleFreq.LOD < 0) {
+            qstar = 0.0;
+            LOD = -1 * alleleFreq.LOD;
+        }else{
+            qstar = alleleFreq.qstar;
+            LOD = alleleFreq.LOD;
+        }                              
+
+        //long confident_calls_at_hapmap_chip_sites = 0;
+        if (LOD >= LOD_cutoff && has_hapmap_chip_genotype) {
+            //confident_calls_at_hapmap_chip_sites++;
+
+            // convert hapmap call to mixture of ref/nonref
+            String hapmap_genotype = hapmap_chip_genotype.getFeature();
+            long refs=0, alts=0;
+            double hapmap_q;
+
+            for (char c : hapmap_genotype.toCharArray()) {
+                if (c == alleleFreq.ref) { refs++; }
+                if (c == alleleFreq.alt) { alts++; }
+            }
+
+            if (refs+alts > 0) {
+                hapmap_q = (float)alts / (refs+alts);
+            }else{
+                hapmap_q = -1;
+            }
+
+            // Hapmap debug info
+            System.out.format("HAPMAP %.2f %.2f %.2f ", hapmap_q, qstar, LOD);
+            String called_genotype = alleleFreq.asString();
+            System.out.format("%s %s %c %c %.2f", hapmap_genotype, called_genotype, alleleFreq.ref, alleleFreq.alt, alleleFreq.LOD);
+
+            // Calculate genotyping performance - did we get the correct genotype of the N+1 choices?
+            if (hapmap_q != -1 && hapmap_q == qstar) {
+                hapmap_genotype_correct++;
+                System.out.print("\n");
+            }else{
+                hapmap_genotype_incorrect++;
+                System.out.printf(" INCORRECT GENOTYPE Bases: %s\n", AlleleFrequencyWalker.getBases(context));
+                AlleleFrequencyWalker.print_base_qual_matrix(AlleleFrequencyWalker.getOneBaseQuals(context));
+            }
+
+            // Now calculate ref / var performance - did we correctly classify the site as
+            // reference or variant without regard to genotype; i.e. het/hom "miscalls" don't matter here
+            boolean hapmap_var = hapmap_q != 0.0;
+            boolean called_var = qstar != 0.0;
+            if (hapmap_var != called_var) {
+                hapmap_refvar_incorrect++;
+            }else{
+                hapmap_refvar_correct++;
+            }
+
+        }
+
+        /*String hapmap_genotype = hapmap_chip_genotype.getFeature();
+        String called_genotype = alleleFreq.asString();
+        System.out.format("HAPMAP %s %s %.2f\n", hapmap_genotype, called_genotype, alleleFreq.LOD);
+        // There is a hapmap site here, is it correct?
+        if (hapmap_genotype.equals(called_genotype))
+        {
+            hapmap_tp++;
+        } else {
+            hapmap_fp++;
+        } */
+
         return alleleFreq;
     }
 
@@ -97,10 +160,18 @@ public class AlleleFrequencyMetricsWalker extends BasicLociWalker<AlleleFrequenc
         {
 	        System.out.printf("METRICS Number of variants                    : %d (%.2f%%) (1/%d)\n", num_variants, 100.0 * (float)num_variants / (float)num_loci_confident, num_loci_confident / num_variants);
             System.out.printf("METRICS Fraction of variant sites in dbSNP    : %.2f%%\n", 100.0 * (float)dbsnp_hits / (float)num_variants);
-            System.out.printf("METRICS Number of variants with hapmap calls : %d\n", hapmap_tp + hapmap_fp);
+            System.out.printf("METRICS -------------------------------------------------\n");
-            System.out.printf("METRICS Hapmap true positives                : %d\n", hapmap_tp);
+            System.out.printf("METRICS        -- Hapmap Genotyping performance --\n");
-            System.out.printf("METRICS Hapmap false positives               : %d\n", hapmap_fp);
+            System.out.printf("METRICS Num. conf. calls at Hapmap chip sites : %d\n", hapmap_genotype_correct + hapmap_genotype_incorrect);
-            System.out.printf("METRICS Hapmap precision                     : %.2f%%\n", 100.0 * (float)hapmap_tp / (float)(hapmap_tp + hapmap_fp));
+            System.out.printf("METRICS Conf. calls at chip sites correct     : %d\n", hapmap_genotype_correct);
+            System.out.printf("METRICS Conf. calls at chip sites incorrect   : %d\n", hapmap_genotype_incorrect);
+            System.out.printf("METRICS %% of confident calls that are correct : %.2f%%\n", 100.0 * (float) hapmap_genotype_correct / (float)(hapmap_genotype_correct + hapmap_genotype_incorrect));
+            System.out.printf("METRICS -------------------------------------------------\n");
+            System.out.printf("METRICS    -- Hapmap Reference/Variant performance --\n");
+            System.out.printf("METRICS Num. conf. calls at Hapmap chip sites : %d\n", hapmap_refvar_correct + hapmap_refvar_incorrect);
+            System.out.printf("METRICS Conf. calls at chip sites correct     : %d\n", hapmap_refvar_correct);
+            System.out.printf("METRICS Conf. calls at chip sites incorrect   : %d\n", hapmap_refvar_incorrect);
+            System.out.printf("METRICS %% of confident calls that are correct : %.2f%%\n", 100.0 * (float) hapmap_refvar_correct / (float)(hapmap_refvar_correct + hapmap_refvar_incorrect));
         }
         System.out.println();
     }
@@ -118,17 +189,8 @@ public class AlleleFrequencyMetricsWalker extends BasicLociWalker<AlleleFrequenc
 
     public String reduce(AlleleFrequencyEstimate alleleFreq, String sum) 
     {
-        if ((alleleFreq.LOD >= 5) || (alleleFreq.LOD <= -5))
+        // Print RESULT data for confident calls
-        {
+        if ((alleleFreq.LOD >= 5) || (alleleFreq.LOD <= -5)) { System.out.print(alleleFreq.asTabularString()); }
-	        System.out.print(String.format("RESULT %s %c %c %f %f %f %d\n", 
-	                                        alleleFreq.location,
-	                                        alleleFreq.ref, 
-	                                        alleleFreq.alt, 
-	                                        alleleFreq.qhat, 
-	                                        alleleFreq.qstar, 
-	                                        alleleFreq.LOD, 
-	                                        alleleFreq.depth));
-        }
 
         if (this.num_loci_total % 10000 == 0) { printMetrics(); }
 

java/src/org/broadinstitute/sting/playground/gatk/walkers/AlleleFrequencyWalker.java

@@ -12,47 +12,26 @@ import java.util.Arrays;
 
 public class AlleleFrequencyWalker extends BasicLociWalker<AlleleFrequencyEstimate, Integer> {
 
-    public AlleleFrequencyEstimate map(List<ReferenceOrderedDatum> rodData, char ref, LocusContext context) {
 
-        // Convert context data into bases and 4-base quals
+    public AlleleFrequencyEstimate map(List<ReferenceOrderedDatum> rodData, char ref, LocusContext context) {
 
         // Set number of chromosomes, N, to 2 for now
         int N = 2;
 
-        // Convert bases to CharArray
+        // Convert context data into bases and 4-base quals
-        int numReads = context.getReads().size(); //numReads();
+        String bases = getBases(context);
-        byte[] bases = new byte[numReads];
+        double quals[][] = getOneBaseQuals(context);
-        String base_string = "";
-        double[][] quals = new double[numReads][4];
-        int refnum = nuc2num[ref];
-
-        List<SAMRecord> reads = context.getReads();
-        List<Integer> offsets = context.getOffsets();
-        for (int i =0; i < numReads; i++ ) {
-            SAMRecord read = reads.get(i);
-            int offset = offsets.get(i);
-
-            bases[i] = read.getReadBases()[offset];
-            base_string += read.getReadString().charAt(offset);
-            int callednum = nuc2num[bases[i]];
-            quals[i][callednum] = 1 - Math.pow(10.0, (double)read.getBaseQualities()[offset] / -10.0);
-
-            // Set all nonref qual scores to their share of the remaining probality not "used" by the reference base's qual
-            double nonref_quals = (1.0 - quals[i][callednum]) / 3;
-            for (int b=0; b<4; b++)
-                if (b != callednum)
-                    quals[i][b] = nonref_quals;
-        }
 
         // Count bases
         int[] base_counts = new int[4];
-        for (byte b : bases)
+        for (byte b : bases.getBytes())
             base_counts[nuc2num[b]]++;
 
         // Find alternate allele - 2nd most frequent non-ref allele
         // (maybe we should check for ties and eval both or check most common including quality scores)
         int altnum = -1; // start with first base numerical identity (0,1,2,3)
         double altcount = -1; // start with first base count
+        int refnum = nuc2num[ref];
 
         for (int b=0; b<4; b++) {
             if ((b != refnum) && (base_counts[b] > altcount)) {
@@ -61,7 +40,7 @@ public class AlleleFrequencyWalker extends BasicLociWalker<AlleleFrequencyEstima
         }
         assert(altnum != -1);
 
-        AlleleFrequencyEstimate alleleFreq = AlleleFrequencyEstimator(context.getLocation().toString(), N, bases, quals, refnum, altnum, base_string.length());
+        AlleleFrequencyEstimate alleleFreq = AlleleFrequencyEstimator(context.getLocation().toString(), N, bases.getBytes(), quals, refnum, altnum, bases.length());
 
         // Print dbSNP data if its there
         if (false) {
@@ -76,6 +55,47 @@ public class AlleleFrequencyWalker extends BasicLociWalker<AlleleFrequencyEstima
         return alleleFreq;
     }
 
+    static public String getBases (LocusContext context) {
+        // Convert bases to CharArray
+        int numReads = context.getReads().size(); //numReads();
+        //byte[] bases = new byte[numReads];
+        String base_string = "";
+        //int refnum = nuc2num[ref];
+
+        List<SAMRecord> reads = context.getReads();
+        List<Integer> offsets = context.getOffsets();
+        for (int i =0; i < numReads; i++ ) {
+            SAMRecord read = reads.get(i);
+            int offset = offsets.get(i);
+
+            //bases[i] = read.getReadBases()[offset];
+            base_string += read.getReadString().charAt(offset);
+        }
+        return base_string;
+    }
+
+    static public double[][] getOneBaseQuals (LocusContext context) {
+        int numReads = context.getReads().size(); //numReads();
+        double[][] quals = new double[numReads][4];
+
+        List<SAMRecord> reads = context.getReads();
+        List<Integer> offsets = context.getOffsets();
+        for (int i =0; i < numReads; i++ ) {
+            SAMRecord read = reads.get(i);
+            int offset = offsets.get(i);
+            int callednum = nuc2num[read.getReadBases()[offset]];
+            quals[i][callednum] = 1 - Math.pow(10.0, (double)read.getBaseQualities()[offset] / -10.0);
+
+            // Set all nonref qual scores to their share of the remaining probality not "used" by the reference base's qual
+            double nonref_quals = (1.0 - quals[i][callednum]) / 3;
+            for (int b=0; b<4; b++)
+                if (b != callednum)
+                    quals[i][b] = nonref_quals;
+        }
+
+        return quals;
+    }
+
     public AlleleFrequencyEstimate AlleleFrequencyEstimator(String location, int N, byte[] bases, double[][] quals, int refnum, int altnum, int depth)
     {
 
@@ -124,6 +144,7 @@ public class AlleleFrequencyWalker extends BasicLociWalker<AlleleFrequencyEstima
                     best_pqG = pqG;
                     best_pG  = pG;
                 }
+                //System.out.format("%.2f %.2f %5.2f %5.2f %5.2f %5.2f\n", q, qstar, pDq, pqG, pG, posterior);
             }
         }
 
@@ -230,17 +251,9 @@ public class AlleleFrequencyWalker extends BasicLociWalker<AlleleFrequencyEstima
 
     public Integer reduce(AlleleFrequencyEstimate alleleFreq, Integer sum) 
     {
-        if (alleleFreq.LOD >= 5)
+        // Print RESULT data for confident calls
-        {
+        if ((alleleFreq.LOD >= 5) || (alleleFreq.LOD <= -5)) { System.out.print(alleleFreq.asTabularString()); }
-	        System.out.print(String.format("RESULT %s %c %c %f %f %f %d\n", 
+
-	                                        alleleFreq.location,
-	                                        alleleFreq.ref, 
-	                                        alleleFreq.alt, 
-	                                        alleleFreq.qhat, 
-	                                        alleleFreq.qstar, 
-	                                        alleleFreq.LOD, 
-	                                        alleleFreq.depth));
-        }
         return 0;
     }
 
@@ -267,9 +280,10 @@ public class AlleleFrequencyWalker extends BasicLociWalker<AlleleFrequencyEstima
 
 
 
-    void print_base_qual_matrix(double [][]quals, int numReads) {
+    static void print_base_qual_matrix(double [][]quals) {
         // Print quals for debugging
         System.out.println("Base quality matrix");
+        int numReads = quals.length;
         for (int b=0; b<4; b++) {
             System.out.print(num2nuc[b]);
             for (int i=0; i < numReads; i++){

java/src/org/broadinstitute/sting/playground/utils/AlleleFrequencyEstimate.java

@@ -35,6 +35,17 @@ public class AlleleFrequencyEstimate {
         this.depth = depth;
     }
 
+    public String asTabularString() {
+        return String.format("RESULT %s %c %c %f %f %f %d\n",
+	                                        location,
+	                                        ref,
+	                                        alt,
+	                                        qhat,
+	                                        qstar,
+	                                        LOD,
+	                                        depth);
+    }
+
     public String asString() {
         // Print out the called bases
         // Notes: switched from qhat to qstar because qhat doesn't work at n=1 (1 observed base) where having a single non-ref