533112afe2a2005e80cdb1f82904ea65032d4302
braney
  Sat Oct 2 11:37:34 2021 -0700
split hg/lib into two separate libaries, one only used by the cgis

diff --git src/hg/cgilib/annoGratorGpVar.c src/hg/cgilib/annoGratorGpVar.c
new file mode 100644
index 0000000..ab8f230
--- /dev/null
+++ src/hg/cgilib/annoGratorGpVar.c
@@ -0,0 +1,861 @@
+/* annoGratorGpVar -- integrate pgSNP or VCF with gene pred and make gpFx predictions */
+
+/* Copyright (C) 2014 The Regents of the University of California 
+ * See README in this or parent directory for licensing information. */
+
+#include "annoGratorGpVar.h"
+#include "annoStreamDbPslPlus.h"
+#include "fa.h"
+#include "genbank.h"
+#include "genePred.h"
+#include "hdb.h"
+#include "hgHgvs.h"
+#include "pgSnp.h"
+#include "vcf.h"
+#include "variant.h"
+#include "gpFx.h"
+#include "seqWindow.h"
+#include "trackHub.h"
+#include "twoBit.h"
+#include "variantProjector.h"
+#include "annoGratorQuery.h"
+
+struct annoGratorGpVar
+{
+    struct annoGrator grator;	// external annoGrator/annoStreamer interface
+    struct lm *lm;		// localmem scratch storage
+    struct dyString *dyScratch;	// dyString for local temporary use
+    struct annoGratorGpVarFuncFilter *funcFilter; // Which categories of effect should we output?
+    enum annoGratorOverlap gpVarOverlapRule;	  // Should we set RJFail if no overlap?
+    struct seqWindow *gSeqWin;  // means of fetching genomic sequence for HGVS term generation
+    boolean hgvsMakeG;          // Generate genomic (g.) HGVS terms only if this is set
+    boolean hgvsMakeCN;         // Generate transcript (n. / c.) HGVS terms only if this is set
+    boolean hgvsMakeP;          // Generate protein (p.) HGVS terms only if this is set
+    boolean hgvsAddParensToP;   // Add parentheses around predicted protein changes (strict HGVS)
+    boolean hgvsBreakDelIns;    // Include deleted sequence (not only ins) e.g. delGGinsAT
+    boolean sourceHasFrames;    // True if transcript annotations include exonFrames column
+    boolean sourceIsBigGenePred;// True if transcript annotations are from bigGenePred
+    boolean sourceIsPslPlus;    // True if transcript annotations are PSL+CDS+seq info
+    boolean protLookupInited;   // For looking up protein accessions in genePred-derived HGVS p.
+    char *protLookupTable;      // "
+    struct hash *protLookupHash;// "
+
+    struct variant *(*variantFromRow)(struct annoGratorGpVar *self, struct annoRow *row,
+				      char *refAllele);
+    // Translate row from whatever format it is (pgSnp or VCF) into generic variant.
+    };
+
+
+static char *aggvAutoSqlStringStart =
+"table genePredWithSO"
+"\"genePred with Sequence Ontology annotation\""
+"(";
+
+static char *aggvAutoSqlStringEnd =
+"string  allele;             \"Allele used to predict functional effect\""
+"string  transcript;         \"ID of affected transcript\""
+"uint    soNumber;           \"Sequence Ontology Number \""
+"uint    detailType;         \"gpFx detail type (1=codingChange, 2=nonCodingExon, 3=intron)\""
+"string  detail0;            \"detail column (per detailType) 0\""
+"string  detail1;            \"detail column (per detailType) 1\""
+"string  detail2;            \"detail column (per detailType) 2\""
+"string  detail3;            \"detail column (per detailType) 3\""
+"string  detail4;            \"detail column (per detailType) 4\""
+"string  detail5;            \"detail column (per detailType) 5\""
+"string  detail6;            \"detail column (per detailType) 6\""
+"string  detail7;            \"detail column (per detailType) 7\""
+"string  detail8;            \"detail column (per detailType) 8\""
+"string  detail9;            \"detail column (per detailType) 9\""
+"string  detail10;           \"detail column (per detailType) 10\""
+"string  hgvsG;              \"HGVS g. term\""
+"string  hgvsCN;             \"HGVS c. or n. term\""
+"string  hgvsP;              \"HGVS p. term\""
+")";
+
+struct asObject *annoGpVarAsObj(struct asObject *sourceAsObj)
+// Return asObject describing fields of internal source plus the fields we add here.
+{
+struct dyString *gpPlusGpFx = dyStringCreate("%s", aggvAutoSqlStringStart);
+// Translate each column back into autoSql text for combining with new output fields:
+struct asColumn *col;
+for (col = sourceAsObj->columnList;  col != NULL;  col = col->next)
+    {
+    if (col->fixedSize)
+	dyStringPrintf(gpPlusGpFx, "%s[%d]\t%s;\t\"%s\"",
+		       col->lowType->name, col->fixedSize, col->name, col->comment);
+    else if (col->isArray || col->isList)
+	{
+	if (col->linkedSizeName)
+	    dyStringPrintf(gpPlusGpFx, "%s[%s]\t%s;\t\"%s\"",
+			   col->lowType->name, col->linkedSizeName, col->name, col->comment);
+	else
+	    errAbort("Neither col->fixedSize nor col->linkedSizeName given for "
+		     "asObj %s column '%s'",
+		     sourceAsObj->name, col->name);
+	}
+    else
+	dyStringPrintf(gpPlusGpFx, "%s\t%s;\t\"%s\"", col->lowType->name, col->name, col->comment);
+    }
+dyStringAppend(gpPlusGpFx, aggvAutoSqlStringEnd);
+struct asObject *asObj = asParseText(gpPlusGpFx->string);
+dyStringFree(&gpPlusGpFx);
+return asObj;
+}
+
+static boolean passesFilter(struct annoGratorGpVar *self, struct gpFx *gpFx)
+/* Based on type of effect, should we print this one? */
+{
+struct annoGratorGpVarFuncFilter *filt = self->funcFilter;
+enum soTerm term = gpFx->soNumber;
+if (term == NMD_transcript_variant)
+    // This one gets special treatment because gpFx->detailType might still be codingChange:
+    return filt->nmdTranscript;
+if (filt->intron && (term == intron_variant || term == complex_transcript_variant))
+    return TRUE;
+if (filt->upDownstream && (term == upstream_gene_variant || term == downstream_gene_variant))
+    return TRUE;
+if (filt->exonLoss && (term == exon_loss_variant))
+    return TRUE;
+if ((filt->exonLoss || filt->cdsNonSyn) && term == transcript_ablation)
+    return TRUE;
+if (filt->utr && (term == _5_prime_UTR_variant || term == _3_prime_UTR_variant))
+    return TRUE;
+if (filt->cdsSyn && term == synonymous_variant)
+    return TRUE;
+if (filt->cdsNonSyn && term != synonymous_variant
+    && (gpFx->detailType == codingChange || term == complex_transcript_variant))
+    return TRUE;
+if (filt->nonCodingExon && term == non_coding_transcript_exon_variant)
+    return TRUE;
+if (filt->splice && (term == splice_donor_variant || term == splice_acceptor_variant ||
+		     term == splice_region_variant))
+    return TRUE;
+if (filt->noVariation && term == no_sequence_alteration)
+    return TRUE;
+return FALSE;
+}
+
+static char *blankIfNull(char *input)
+{
+if (input == NULL)
+    return "";
+
+return input;
+}
+
+static char *uintToString(struct lm *lm, uint num)
+{
+char buffer[10];
+
+safef(buffer,sizeof buffer, "%d", num);
+return lmCloneString(lm, buffer);
+}
+
+static void aggvStringifyGpFx(char **words, struct gpFx *effect, struct lm *lm)
+// turn gpFx structure into array of words
+{
+int count = 0;
+
+words[count++] = lmCloneString(lm, effect->gAllele);
+words[count++] = lmCloneString(lm, blankIfNull(effect->transcript));
+words[count++] = uintToString(lm, effect->soNumber);
+words[count++] = uintToString(lm, effect->detailType);
+int gpFxNumCols = 4;
+
+if (effect->detailType == intron)
+    {
+    words[count++] = uintToString(lm, effect->details.intron.intronNumber);
+    words[count++] = uintToString(lm, effect->details.intron.intronCount);
+    }
+else if (effect->detailType == nonCodingExon)
+    {
+    words[count++] = uintToString(lm, effect->details.nonCodingExon.exonNumber);
+    words[count++] = uintToString(lm, effect->details.nonCodingExon.exonCount);
+    words[count++] = uintToString(lm, effect->details.nonCodingExon.cDnaPosition);
+    words[count++] = lmCloneString(lm, effect->details.nonCodingExon.txRef);
+    words[count++] = lmCloneString(lm, effect->details.nonCodingExon.txAlt);
+    }
+else if (effect->detailType == codingChange)
+    {
+    struct codingChange *cc = &(effect->details.codingChange);
+    words[count++] = uintToString(lm, cc->exonNumber);
+    words[count++] = uintToString(lm, cc->exonCount);
+    words[count++] = uintToString(lm, cc->cDnaPosition);
+    words[count++] = lmCloneString(lm, cc->txRef);
+    words[count++] = lmCloneString(lm, cc->txAlt);
+    words[count++] = uintToString(lm, cc->cdsPosition);
+    words[count++] = uintToString(lm, cc->pepPosition);
+    words[count++] = strUpper(lmCloneString(lm, blankIfNull(cc->aaOld)));
+    words[count++] = strUpper(lmCloneString(lm, blankIfNull(cc->aaNew)));
+    words[count++] = strUpper(lmCloneString(lm, blankIfNull(cc->codonOld)));
+    words[count++] = strUpper(lmCloneString(lm, blankIfNull(cc->codonNew)));
+    }
+else if (effect->detailType != none)
+    errAbort("annoGratorGpVar: unknown effect type %d", effect->detailType);
+
+// Add max number of columns added in any if clause above
+gpFxNumCols += 9;
+
+while (count < gpFxNumCols)
+    words[count++] = "";
+}
+
+struct gpFx *annoGratorGpVarGpFxFromRow(struct annoStreamer *sSelf, struct annoRow *row,
+					struct lm *lm)
+// Turn the string array representation back into a real gpFx.
+// I know this is inefficient and am thinking about a better way.
+{
+if (row == NULL)
+    return NULL;
+struct gpFx *effect;
+lmAllocVar(lm, effect);
+struct annoGrator *gSelf = (struct annoGrator *)sSelf;
+// get gpFx words which follow internal source's words:
+char **words = (char **)(row->data);
+int count = gSelf->mySource->numCols;
+
+effect->gAllele = lmCloneString(lm, words[count++]);
+effect->transcript = lmCloneString(lm, words[count++]);
+effect->soNumber = atol(words[count++]);
+effect->detailType = atol(words[count++]);
+
+if (effect->detailType == intron)
+    {
+    effect->details.intron.intronNumber = atol(words[count++]);
+    effect->details.intron.intronCount = atol(words[count++]);
+    }
+else if (effect->detailType == nonCodingExon)
+    {
+    effect->details.nonCodingExon.exonNumber = atol(words[count++]);
+    effect->details.nonCodingExon.exonCount = atol(words[count++]);
+    effect->details.nonCodingExon.cDnaPosition = atol(words[count++]);
+    effect->details.nonCodingExon.txRef = lmCloneString(lm, words[count++]);
+    effect->details.nonCodingExon.txAlt = lmCloneString(lm, words[count++]);
+    }
+else if (effect->detailType == codingChange)
+    {
+    effect->details.codingChange.exonNumber = atol(words[count++]);
+    effect->details.codingChange.exonCount = atol(words[count++]);
+    effect->details.codingChange.cDnaPosition = atol(words[count++]);
+    effect->details.codingChange.txRef = lmCloneString(lm, words[count++]);
+    effect->details.codingChange.txAlt = lmCloneString(lm, words[count++]);
+    effect->details.codingChange.cdsPosition = atol(words[count++]);
+    effect->details.codingChange.pepPosition = atol(words[count++]);
+    effect->details.codingChange.aaOld = lmCloneString(lm, words[count++]);
+    effect->details.codingChange.aaNew = lmCloneString(lm, words[count++]);
+    effect->details.codingChange.codonOld = lmCloneString(lm, words[count++]);
+    effect->details.codingChange.codonNew = lmCloneString(lm, words[count++]);
+    }
+else if (effect->detailType != none)
+    errAbort("annoGratorGpVar: unknown effect type %d", effect->detailType);
+return effect;
+}
+
+// Container for optional HGVS terms
+struct hgvsTerms
+    {
+    char *hgvsG;     // HGVS g. term or NULL
+    char *hgvsCN;    // HGVS c. or n. term or NULL
+    char *hgvsP;     // HGVS p. term or NULL
+    };
+
+static  void hgvsTermsFree(struct hgvsTerms *hgvs)
+/* Free struct hgvsTerms and its members. */
+{
+if (hgvs)
+    {
+    freez(&hgvs->hgvsG);
+    freez(&hgvs->hgvsCN);
+    freez(&hgvs->hgvsP);
+    freeMem(hgvs);
+    }
+}
+
+static struct annoRow *aggvEffectToRow(struct annoGratorGpVar *self, struct gpFx *effect,
+				       struct annoRow *rowIn, struct hgvsTerms *hgvs,
+                                       struct lm *callerLm)
+// convert a single genePred annoRow and gpFx record to an augmented genePred annoRow;
+{
+struct annoGrator *gSelf = &(self->grator);
+struct annoStreamer *sSelf = &(gSelf->streamer);
+assert(sSelf->numCols > gSelf->mySource->numCols);
+
+char **wordsOut;
+lmAllocArray(self->lm, wordsOut, sSelf->numCols);
+
+// copy the genePred fields over
+int gpColCount = gSelf->mySource->numCols;
+char **wordsIn = (char **)rowIn->data;
+memcpy(wordsOut, wordsIn, sizeof(char *) * gpColCount);
+
+// stringify the gpFx structure 
+aggvStringifyGpFx(&wordsOut[gpColCount], effect, callerLm);
+
+// Final columns: optional HGVS terms
+wordsOut[sSelf->numCols-3] = lmCloneString(callerLm, hgvs && hgvs->hgvsG ? hgvs->hgvsG : "");
+wordsOut[sSelf->numCols-2] = lmCloneString(callerLm, hgvs && hgvs->hgvsCN ? hgvs->hgvsCN : "");
+wordsOut[sSelf->numCols-1] = lmCloneString(callerLm, hgvs && hgvs->hgvsP ? hgvs->hgvsP : "");
+
+return annoRowFromStringArray(rowIn->chrom, rowIn->start, rowIn->end, rowIn->rightJoinFail,
+			      wordsOut, sSelf->numCols, callerLm);
+}
+
+struct dnaSeq *genePredToGenomicSequence(struct genePred *pred, struct annoAssembly *aa,
+					 struct lm *lm)
+/* Return concatenated genomic sequence of exons of pred. */
+{
+int txLen = 0;
+int i;
+for (i=0; i < pred->exonCount; i++)
+    txLen += (pred->exonEnds[i] - pred->exonStarts[i]);
+char *seq = lmAlloc(lm, txLen + 1);
+int offset = 0;
+for (i=0; i < pred->exonCount; i++)
+    {
+    int exonStart = pred->exonStarts[i];
+    int exonEnd = pred->exonEnds[i];
+    annoAssemblyGetSeq(aa, pred->chrom, exonStart, exonEnd, seq+offset, txLen+1-offset);
+    offset += (exonEnd - exonStart);
+    }
+if(pred->strand[0] == '-')
+    reverseComplement(seq, txLen);
+struct dnaSeq *txSeq = NULL;
+lmAllocVar(lm, txSeq);
+txSeq->name = lmCloneString(lm, pred->name);
+txSeq->dna = seq;
+txSeq->size = txLen;
+return txSeq;
+}
+
+struct dnaSeq *translateTx(struct dnaSeq *txSeq, struct genbankCds *cds)
+/* Translate txSeq into protein sequence, including 'X' for stop codon if present. */
+{
+struct dnaSeq *protSeq = translateSeq(txSeq, cds->start, FALSE);
+aaSeqZToX(protSeq);
+return protSeq;
+}
+
+static struct udcFile *getCachedUdcFile(char *fileOrUrl)
+/* Return an open UDC file handle for fileOrUrl; cache open connections.  errAbort if can't open. */
+{
+static struct hash *hash = NULL;
+if (hash == NULL)
+    hash = hashNew(0);
+struct udcFile *udcf = hashFindVal(hash, fileOrUrl);
+if (udcf == NULL)
+    {
+    char *realFileOrUrl = hReplaceGbdb(fileOrUrl);
+    udcf = udcFileOpen(realFileOrUrl, NULL);
+    hashAdd(hash, fileOrUrl, udcf);
+    freeMem(realFileOrUrl);
+    }
+return udcf;
+}
+
+static struct dnaSeq *getCachedSeq(char *fileOrUrl, off_t offset, size_t size, boolean isDna)
+/* Parse FASTA in file fileOrUrl at offset and return a cached dnaSeq. */
+{
+static struct hash *hash = NULL;
+if (hash == NULL)
+    hash = hashNew(0);
+char key[4096];
+safef(key, sizeof(key), "%s_%lld_%lld", fileOrUrl, (long long)offset, (long long)size);
+struct dnaSeq *seq = hashFindVal(hash, key);
+if (seq == NULL)
+    {
+    char *buf = needMem(size+1);
+    struct udcFile *udcf = getCachedUdcFile(fileOrUrl);
+    udcSeek(udcf, offset);
+    udcRead(udcf, buf, size);
+    seq = faSeqFromMemText(buf, isDna);
+    toUpperN(seq->dna, seq->size);
+    hashAdd(hash, key, seq);
+    }
+return seq;
+}
+
+struct dnaSeq *getProtSeq(struct annoGratorGpVar *self, char *protAcc,
+                          struct dnaSeq *txSeq, struct genbankCds *cds)
+/* If protAcc is NULL, translate txSeq+cds; else look up protAcc. */
+{
+static struct hash *hash = NULL;
+if (hash == NULL)
+    hash = hashNew(0);
+struct dnaSeq *accSeq = NULL;
+if (isEmpty(protAcc))
+    {
+    accSeq = hashFindVal(hash, txSeq->name);
+    if (accSeq == NULL)
+        {
+        accSeq = translateTx(txSeq, cds);
+        hashAdd(hash, txSeq->name, accSeq);
+        }
+    }
+else
+    {
+    accSeq = hashFindVal(hash, protAcc);
+    char *db = self->grator.streamer.assembly->name;
+    char query[1024];
+    struct sqlConnection *conn = hAllocConn(db);
+    //#*** Using presence or absence of dot-version is an ugly hack, but it will do for now;
+    //#*** otherwise there's new config to pass in & store...
+    if (strchr(protAcc, '.'))
+        {
+        // Use ncbiRefSeqPepTable
+        sqlSafef(query, sizeof(query), "select seq from ncbiRefSeqPepTable where name = '%s'",
+                 protAcc);
+        char *seq = sqlQuickString(conn, query);
+        if (seq)
+            accSeq = newDnaSeq(seq, strlen(seq), protAcc);
+        }
+    else
+        {
+        // Get GenBank versioned acc and seq
+        sqlSafef(query, sizeof(query),
+                 "select path,file_offset,file_size from %s,%s "
+                 "where acc = '%s' and gbExtFile.id = gbExtFile",
+                 gbSeqTable, gbExtFileTable, protAcc);
+        char **row;
+        struct sqlResult *sr = sqlGetResult(conn, query);
+        if ((row = sqlNextRow(sr)) != NULL)
+            accSeq = getCachedSeq(row[0], atoll(row[1]), atoll(row[2]), FALSE);
+        sqlFreeResult(&sr);
+        }
+    if (accSeq == NULL)
+        // Store a dnaSeq with NULL name and seq so we don't waste time sql querying this again.
+        accSeq = newDnaSeq(NULL, 0, NULL);
+    hashAdd(hash, protAcc, accSeq);
+    hFreeConn(&conn);
+    }
+return (accSeq->name == NULL) ? NULL : accSeq;
+}
+
+static struct hgvsTerms *getHgvsTerms(struct annoGratorGpVar *self, char *chromAcc,
+                                      struct psl *psl, struct genbankCds *cds,
+                                      struct dnaSeq *txSeq, struct dnaSeq *protSeq,
+                                      struct vpTx *vpTx, struct vpPep *vpPep)
+/* Return HGVS terms for a variant allele projected onto the genome. */
+{
+struct hgvsTerms *hgvs;
+AllocVar(hgvs);
+struct bed3 variantBed;
+variantBed.chrom = psl->tName;
+variantBed.chromStart = min(vpTx->start.gOffset, vpTx->end.gOffset);
+variantBed.chromEnd = max(vpTx->start.gOffset, vpTx->end.gOffset);
+if (self->hgvsMakeG)
+    {
+    int gAltLen = strlen(vpTx->gAlt);
+    char gAlt[gAltLen+1];
+    safecpy(gAlt, sizeof(gAlt), vpTx->gAlt);
+    if (pslQStrand(psl) == '-')
+        reverseComplement(gAlt, gAltLen);
+    hgvs->hgvsG = hgvsGFromVariant(self->gSeqWin, &variantBed, gAlt, chromAcc,
+                                   self->hgvsBreakDelIns);
+    }
+if ((self->hgvsMakeCN || self->hgvsMakeP) && txSeq)
+    {
+    if (cds->start != cds->end && cds->start >= 0)
+        {
+        if (self->hgvsMakeCN)
+            hgvs->hgvsCN = hgvsCFromVpTx(vpTx, self->gSeqWin, psl, cds, txSeq,
+                                         self->hgvsBreakDelIns);
+        if (self->hgvsMakeP)
+            {
+            hgvs->hgvsP = hgvsPFromVpPep(vpPep, protSeq, self->hgvsAddParensToP);
+            }
+        }
+    else if (self->hgvsMakeCN)
+        hgvs->hgvsCN = hgvsNFromVpTx(vpTx, self->gSeqWin, psl, txSeq,
+                                     self->hgvsBreakDelIns);
+    }
+return hgvs;
+}
+
+
+static struct annoRow *aggvPredict(struct annoGratorGpVar *self, struct variant *variant,
+                                   struct psl *psl, struct genbankCds *cds,
+                                   struct dnaSeq *txSeq, struct dnaSeq *protSeq,
+                                   struct annoRow *inRow, struct lm *callerLm)
+// Project variant through psl onto transcript and predict functional effects.
+{
+struct annoRow *rowList = NULL;
+char *db = self->grator.streamer.assembly->name;
+char *chromAcc = self->hgvsMakeG ? hRefSeqAccForChrom(db, variant->chrom) : NULL;
+struct bed3 *variantBed = (struct bed3 *)variant;
+vpExpandIndelGaps(psl, self->gSeqWin, txSeq);
+struct allele *allele;
+for (allele = variant->alleles;  allele != NULL;  allele = allele->next)
+    {
+    char *alt = allele->sequence;
+    struct vpTx *vpTx = vpGenomicToTranscript(self->gSeqWin, variantBed, alt, psl, txSeq);
+    if (!allele->isReference || vpTx->genomeMismatch)
+        {
+        struct vpPep *vpPep = vpTranscriptToProtein(vpTx, cds, txSeq, protSeq);
+        struct hgvsTerms *hgvs = getHgvsTerms(self, chromAcc, psl, cds, txSeq, protSeq,
+                                              vpTx, vpPep);
+        struct gpFx *fxList = vpTranscriptToGpFx(vpTx, psl, cds, txSeq, vpPep, protSeq, self->lm);
+        struct annoRow *rows = NULL;
+        struct gpFx *fx;
+        for(fx = fxList;  fx != NULL; fx = fx->next)
+            {
+            // Intergenic variants never pass through here so we don't have to filter them out
+            // here if self->funcFilter is null.
+            if (self->funcFilter == NULL || passesFilter(self, fx))
+                {
+                // Restore the original variant's alt allele
+                fx->gAllele = lmCloneString(self->lm, allele->sequence);
+                slAddHead(&rows, aggvEffectToRow(self, fx, inRow, hgvs, callerLm));
+                }
+            }
+        slReverse(&rows);
+        rowList = slCat(rows, rowList);
+        hgvsTermsFree(hgvs);
+        vpPepFree(&vpPep);
+        }
+    vpTxFree(&vpTx);
+    }
+freeMem(chromAcc);
+return rowList;
+}
+
+static void lookupProtAcc(struct annoGratorGpVar *self, struct dnaSeq *protSeq)
+/* For Gencode, try to find the ENSP* ID associated with the ENST* ID. */
+{
+char *db = self->grator.streamer.assembly->name;
+if (! self->protLookupInited)
+    {
+    char *sourceName = self->grator.streamer.name;
+    if (strstr(sourceName, "wgEncodeGencode"))
+        {
+        char *version = strrchr(sourceName, 'V');
+        if (version)
+            {
+            if (!trackHubDatabase(db))
+                {
+                char attrsTable[512];
+                safef(attrsTable, sizeof(attrsTable), "wgEncodeGencodeAttrs%s", version);
+                if (hTableExists(db, attrsTable) && hHasField(db, attrsTable, "proteinId"))
+                    {
+                    self->protLookupHash = hashNew(0);
+                    self->protLookupTable = cloneString(attrsTable);
+                    }
+                }
+            }
+        }
+    self->protLookupInited = TRUE;
+    }
+if (self->protLookupHash != NULL)
+    {
+    char *txAcc = protSeq->name;
+    struct hashEl *hel = hashLookup(self->protLookupHash, txAcc);
+    if (hel == NULL)
+        {
+        struct sqlConnection *conn = hAllocConn(db);
+        char query[2048];
+        sqlSafef(query, sizeof(query), "select proteinId from %s where transcriptId = '%s'",
+                 self->protLookupTable, txAcc);
+        char *protAcc = sqlQuickString(conn, query);
+        hel = hashAdd(self->protLookupHash, txAcc, protAcc);
+        hFreeConn(&conn);
+        }
+    if (hel->val != NULL)
+        {
+        freeMem(protSeq->name);
+        protSeq->name = cloneString(hel->val);
+        }
+    }
+}
+
+static struct annoRow *aggvGenRowsGp(struct annoGratorGpVar *self, struct variant *variant,
+                                     struct genePred *pred, struct annoRow *inRow,
+                                     struct lm *callerLm)
+// put out annoRows for all the gpFx that arise from variant and pred
+{
+struct annoStreamer *sSelf = &(self->grator.streamer);
+struct genbankCds cds;
+genePredToCds(pred, &cds);
+struct dnaSeq *txSeq = genePredToGenomicSequence(pred, sSelf->assembly, self->lm);
+int chromSize = 0;  // unused
+struct psl *psl = genePredToPsl(pred, chromSize, txSeq->size);
+pslRemoveFrameShifts(psl);
+vpExpandIndelGaps(psl, self->gSeqWin, txSeq);
+struct dnaSeq *protSeq = NULL;
+if (cds.end > cds.start)
+    {
+    //#*** what if cds.startComplete is not true??
+    protSeq = translateTx(txSeq, &cds);
+    lookupProtAcc(self, protSeq);
+    }
+struct annoRow *rowList = aggvPredict(self, variant, psl, &cds, txSeq, protSeq, inRow, callerLm);
+pslFree(&psl);
+dnaSeqFree(&protSeq);
+return rowList;
+}
+
+static struct annoRow *aggvGenRowsPsl(struct annoGratorGpVar *self, struct variant *variant,
+                                     struct annoRow *inRow, struct lm *callerLm)
+// put out annoRows for all the gpFx that arise from variant and transcript psl+
+{
+char **ppWords = inRow->data;
+struct psl *psl = pslLoad(ppWords);
+struct genbankCds cds;
+genbankCdsParse(ppWords[PSLPLUS_CDS_IX], &cds);
+struct dnaSeq *txSeq = getCachedSeq(ppWords[PSLPLUS_PATH_IX], atoll(ppWords[PSLPLUS_FILEOFFSET_IX]),
+                                    atoll(ppWords[PSLPLUS_FILESIZE_IX]), TRUE);
+struct dnaSeq *protSeq = getProtSeq(self, ppWords[PSLPLUS_PROTACC_IX], txSeq, &cds);
+struct annoRow *rowList = aggvPredict(self, variant, psl, &cds, txSeq, protSeq, inRow, callerLm);
+pslFree(&psl);
+return rowList;
+}
+
+struct annoRow *aggvGenelessRow(struct annoGratorGpVar *self, struct variant *variant,
+                                boolean rjFail, struct lm *callerLm)
+/* If intergenic variants (no overlapping or nearby genes) are to be included in output,
+ * make an output row with empty genePred and a gpFx that is empty except for soNumber. */
+{
+struct annoGrator *gSelf = &(self->grator);
+struct annoStreamer *sSelf = &(gSelf->streamer);
+char **wordsOut;
+lmAllocArray(self->lm, wordsOut, sSelf->numCols);
+// Add empty strings for genePred string columns:
+int gpColCount = gSelf->mySource->numCols;
+int i;
+for (i = 0;  i < gpColCount;  i++)
+    wordsOut[i] = "";
+struct gpFx *gpFx;
+lmAllocVar(self->lm, gpFx);
+enum soTerm term = hasAltAllele(variant->alleles) ? intergenic_variant : no_sequence_alteration;
+if (term == no_sequence_alteration)
+    gpFx->gAllele = variant->alleles->sequence;
+else
+    gpFx->gAllele = firstAltAllele(variant->alleles);
+if (isAllNt(gpFx->gAllele, strlen(gpFx->gAllele)))
+    touppers(gpFx->gAllele);
+gpFx->soNumber = term;
+gpFx->detailType = none;
+aggvStringifyGpFx(&wordsOut[gpColCount], gpFx, self->lm);
+if (self->hgvsMakeG)
+    {
+    // Add HGVS genomic term
+    struct bed3 *variantBed = (struct bed3 *)variant;
+    char *chromAcc = hRefSeqAccForChrom(sSelf->assembly->name, variant->chrom);
+    char *hgvsG = hgvsGFromVariant(self->gSeqWin, variantBed, gpFx->gAllele, chromAcc,
+                                   self->hgvsBreakDelIns);
+    wordsOut[sSelf->numCols-3] = lmCloneString(callerLm, hgvsG);
+    freeMem(chromAcc);
+    freeMem(hgvsG);
+    }
+return annoRowFromStringArray(variant->chrom, variant->chromStart, variant->chromEnd, rjFail,
+			      wordsOut, sSelf->numCols, callerLm);
+}
+
+static struct variant *variantFromPgSnpTableRow(struct annoGratorGpVar *self, struct annoRow *row,
+                                                char *refAllele)
+/* Translate pgSnp array of words into variant. */
+{
+return variantFromPgSnpAnnoRow(row, refAllele, TRUE, self->lm);
+}
+
+static struct variant *variantFromPgSnpFileRow(struct annoGratorGpVar *self, struct annoRow *row,
+                                               char *refAllele)
+/* Translate pgSnp array of words into variant. */
+{
+return variantFromPgSnpAnnoRow(row, refAllele, FALSE, self->lm);
+}
+
+static struct variant *variantFromVcfRow(struct annoGratorGpVar *self, struct annoRow *row,
+					 char *refAllele)
+/* Translate vcf array of words into variant. */
+{
+return variantFromVcfAnnoRow(row, refAllele, self->lm, self->dyScratch);
+}
+
+static void setVariantFromRow(struct annoGratorGpVar *self, struct annoStreamRows *primaryData)
+/* Depending on autoSql definition of primary source, choose a function to translate
+ * incoming rows into generic variants. */
+{
+if (asObjectsMatch(primaryData->streamer->asObj, pgSnpAsObj()))
+    self->variantFromRow = variantFromPgSnpTableRow;
+else if (asObjectsMatch(primaryData->streamer->asObj, pgSnpFileAsObj()))
+    self->variantFromRow = variantFromPgSnpFileRow;
+else if (asObjectsMatch(primaryData->streamer->asObj, vcfAsObj()))
+    self->variantFromRow = variantFromVcfRow;
+}
+
+struct annoRow *annoGratorGpVarIntegrate(struct annoGrator *gSelf,
+					 struct annoStreamRows *primaryData,
+					 boolean *retRJFilterFailed, struct lm *callerLm)
+// integrate a variant and a genePred, generate as many rows as
+// needed to capture all the changes
+{
+struct annoGratorGpVar *self = (struct annoGratorGpVar *)gSelf;
+struct annoStreamer *sSelf = &(gSelf->streamer);
+lmCleanup(&(self->lm));
+self->lm = lmInit(0);
+if (self->variantFromRow == NULL)
+    setVariantFromRow(self, primaryData);
+// TODO Performance improvement: instead of creating the transcript sequence for each
+// variant that intersects the transcript, cache transcript sequence; possibly
+// an slPair with a concatenation of {chrom, txStart, txEnd, cdsStart, cdsEnd,
+// exonStarts, exonEnds} as the name, and sequence as the val.  When something in
+// the list is no longer in the list of rows from the internal annoGratorIntegrate call,
+// drop it.
+// BETTER YET: make a callback for gpFx to get CDS sequence only when it needs it.
+struct annoRow *primaryRow = primaryData->rowList;
+int refAlBufSize = primaryRow->end - primaryRow->start + 1;
+char refAllele[refAlBufSize];
+annoAssemblyGetSeq(sSelf->assembly, primaryRow->chrom, primaryRow->start, primaryRow->end,
+		   refAllele, sizeof(refAllele));
+struct variant *variant = self->variantFromRow(self, primaryRow, refAllele);
+
+// Temporarily tweak primaryRow's start and end to find upstream/downstream overlap:
+int pStart = primaryRow->start, pEnd = primaryRow->end;
+if (primaryRow->start <= GPRANGE)
+    primaryRow->start = 0;
+else
+    primaryRow->start -= GPRANGE;
+primaryRow->end += GPRANGE;
+struct annoRow *rows = annoGratorIntegrate(gSelf, primaryData, retRJFilterFailed, self->lm);
+primaryRow->start = pStart;
+primaryRow->end = pEnd;
+
+if (rows == NULL)
+    {
+    // No genePreds means that the primary variant is intergenic.
+    if ((self->funcFilter == NULL || self->funcFilter->intergenic))
+        return aggvGenelessRow(self, variant, *retRJFilterFailed, callerLm);
+    else if (retRJFilterFailed && self->gpVarOverlapRule == agoMustOverlap)
+	*retRJFilterFailed = TRUE;
+    return NULL;
+    }
+if (retRJFilterFailed && *retRJFilterFailed)
+    return NULL;
+
+struct annoRow *outRows = NULL;
+
+for(; rows; rows = rows->next)
+    {
+    struct annoRow *outRow = NULL;
+    if (self->sourceIsPslPlus)
+        {
+        outRow = aggvGenRowsPsl(self, variant, rows, callerLm);
+        }
+    else
+        {
+        struct genePred *gp;
+        if (self->sourceIsBigGenePred)
+            gp = (struct genePred *)genePredFromBigGenePredRow(rows->data);
+        else
+            {
+            // work around genePredLoad's trashing its input
+            char **inWords = rows->data;
+            char *saveExonStarts = lmCloneString(self->lm, inWords[8]);
+            char *saveExonEnds = lmCloneString(self->lm, inWords[9]);
+            gp = self->sourceHasFrames ? genePredExtLoad(inWords, GENEPREDX_NUM_COLS) :
+                                         genePredLoad(inWords);
+            inWords[8] = saveExonStarts;
+            inWords[9] = saveExonEnds;
+            }
+        outRow = aggvGenRowsGp(self, variant, gp, rows, callerLm);
+        genePredFree(&gp);
+        }
+    if (outRow != NULL)
+        {
+        slReverse(&outRow);
+        outRows = slCat(outRow, outRows);
+        }
+    }
+slReverse(&outRows);
+// If all rows failed the filter, and we must overlap, set *retRJFilterFailed.
+if (outRows == NULL && retRJFilterFailed && self->gpVarOverlapRule == agoMustOverlap)
+    *retRJFilterFailed = TRUE;
+return outRows;
+}
+
+static void aggvSetOverlapRule(struct annoGrator *gSelf, enum annoGratorOverlap rule)
+/* We need an overlap rule that is independent of the genePred integration because
+ * if we're including intergenic variants, then we return a row even though no genePred
+ * rows overlap. */
+{
+struct annoGratorGpVar *self = (struct annoGratorGpVar *)gSelf;
+self->gpVarOverlapRule = rule;
+// For mustOverlap, if we're including intergenic variants in output, don't let simple
+// integration set retRjFilterFailed:
+if (rule == agoMustOverlap && self->funcFilter != NULL && self->funcFilter->intergenic)
+    gSelf->overlapRule = agoNoConstraint;
+else
+    gSelf->overlapRule = rule;
+}
+
+void aggvClose(struct annoStreamer **pSSelf)
+/* Close out localmem and all the usual annoGrator stuff. */
+{
+if (*pSSelf != NULL)
+    {
+    struct annoGratorGpVar *self = (struct annoGratorGpVar *)(*pSSelf);
+    lmCleanup(&(self->lm));
+    dyStringFree(&(self->dyScratch));
+    freez(&self->protLookupTable);
+    hashFree(&self->protLookupHash);
+    annoGratorClose(pSSelf);
+    }
+}
+
+struct annoGrator *annoGratorGpVarNew(struct annoStreamer *mySource)
+/* Make a subclass of annoGrator that combines genePreds from mySource with
+ * pgSnp rows from primary source to predict functional effects of variants
+ * on genes.
+ * mySource becomes property of the new annoGrator (don't close it, close annoGrator). */
+{
+struct annoGratorGpVar *self;
+AllocVar(self);
+struct annoGrator *gSelf = &(self->grator);
+annoGratorInit(gSelf, mySource);
+struct annoStreamer *sSelf = &(gSelf->streamer);
+// We add columns beyond what comes from mySource, so update our public-facing asObject:
+annoGratorSetAutoSqlObject(sSelf, annoGpVarAsObj(mySource->asObj));
+gSelf->setOverlapRule = aggvSetOverlapRule;
+sSelf->close = aggvClose;
+// integrate by adding gpFx fields
+gSelf->integrate = annoGratorGpVarIntegrate;
+self->dyScratch = dyStringNew(0);
+self->sourceHasFrames = (asColumnFindIx(mySource->asObj->columnList, "exonFrames") >= 0);
+self->sourceIsBigGenePred = (asColumnFindIx(mySource->asObj->columnList, "chromStart") >= 0);
+self->sourceIsPslPlus = (asColumnFindIx(mySource->asObj->columnList, "tStart") >= 0);
+char *db = sSelf->assembly->name;
+self->gSeqWin = chromSeqWindowNew(db, NULL, 0, 0);
+
+return gSelf;
+}
+
+void annoGratorGpVarSetFuncFilter(struct annoGrator *gSelf,
+				  struct annoGratorGpVarFuncFilter *funcFilter)
+/* If funcFilter is non-NULL, it specifies which functional categories
+ * to include in output; if NULL, by default intergenic variants are excluded and
+ * all other categories are included.
+ * NOTE: This calls gSelf->setOverlapRule() with the currently set overlap rule because
+ * overlapRule is affected by filter settings.  */
+{
+struct annoGratorGpVar *self = (struct annoGratorGpVar *)gSelf;
+freez(&self->funcFilter);
+if (funcFilter != NULL)
+    self->funcFilter = CloneVar(funcFilter);
+// Since our overlapRule behavior depends on filter settings, reevaluate:
+gSelf->setOverlapRule(gSelf, self->gpVarOverlapRule);
+}
+
+void annoGratorGpVarSetHgvsOutOptions(struct annoGrator *gSelf, uint hgvsOutOptions)
+/* Import the HGVS output options described in hgHgvs.h */
+{
+struct annoGratorGpVar *self = (struct annoGratorGpVar *)gSelf;
+if (hgvsOutOptions & HGVS_OUT_G)
+    self->hgvsMakeG = TRUE;
+if (hgvsOutOptions & HGVS_OUT_CN)
+    self->hgvsMakeCN = TRUE;
+if (hgvsOutOptions & HGVS_OUT_P)
+    self->hgvsMakeP = TRUE;
+if (hgvsOutOptions & HGVS_OUT_P_ADD_PARENS)
+    self->hgvsAddParensToP = TRUE;
+if (hgvsOutOptions & HGVS_OUT_BREAK_DELINS)
+    self->hgvsBreakDelIns = TRUE;
+}