public VepPredictionParser(VCFHeader header, String tag) {
this.tag = tag;
VCFInfoHeaderLine info = header.getInfoHeaderLine(tag);
if (info == null || info.getDescription() == null) {
System.err.println("NO " + tag + " found in header");
return;
}
String description = info.getDescription();
String chunck = " Format:";
int i = description.indexOf(chunck);
if (i == -1) {
System.err.println("Cannot find " + chunck + " in " + description);
return;
}
description =
description.substring(i + chunck.length()).replaceAll("[ \'\\.\\(\\)]+", "").trim();
String tokens[] = pipe.split(description);
for (i = 0; i < tokens.length; ++i) {
if (tokens[i].isEmpty()) continue;
COLS col = null;
for (COLS c : COLS.values()) {
if (c.name().equalsIgnoreCase(tokens[i])) {
col = c;
}
}
if (col == null) {
System.err.println("Undefined VEP tag " + tokens[i]);
}
col2col.put(col, i);
}
}
/**
* Gets the header fields from all VCF rods input by the user
*
* @param toolkit GATK engine
* @param rodNames names of rods to use, or null if we should use all possible ones
* @return a set of all fields
*/
public static Set<VCFHeaderLine> getHeaderFields(
GenomeAnalysisEngine toolkit, Collection<String> rodNames) {
// keep a map of sample name to occurrences encountered
TreeSet<VCFHeaderLine> fields = new TreeSet<VCFHeaderLine>();
// iterate to get all of the sample names
List<ReferenceOrderedDataSource> dataSources = toolkit.getRodDataSources();
for (ReferenceOrderedDataSource source : dataSources) {
// ignore the rod if it's not in our list
if (rodNames != null && !rodNames.contains(source.getName())) continue;
if (source.getRecordType().equals(VariantContext.class)) {
VCFHeader header = (VCFHeader) source.getHeader();
if (header != null) fields.addAll(header.getMetaDataInSortedOrder());
}
}
return fields;
}
private void read(InputStream in, String filename) throws IOException {
// Pattern comma=Pattern.compile("[,]");
Pattern pipe = Pattern.compile("[\\|]");
Pattern amp = Pattern.compile("&");
out.println("insert into FILE" + SUFFIX + "(filename) values (" + quote(filename) + ");");
VcfIterator r = new VcfIterator(in);
VCFHeader header = r.getHeader();
String csqColumns[] = null;
VCFInfoHeaderLine infoHeader = header.getInfoHeaderLine("CSQ");
if (infoHeader != null && this.USE_VEP) {
LOG.info("parsing VEP " + infoHeader.getDescription());
final String formatStr = "Format: ";
int i = infoHeader.getDescription().indexOf(formatStr);
if (i != -1) {
csqColumns =
pipe.split(infoHeader.getDescription().substring(i + formatStr.length()).trim());
LOG.debug(Arrays.asList(csqColumns));
} else {
LOG.error("Cannot parse " + infoHeader.getDescription());
}
}
String snpEffColumns[] = null;
infoHeader = header.getInfoHeaderLine("EFF");
if (infoHeader != null && this.USE_SNPEFF) {
LOG.info("parsing EFF " + infoHeader.getDescription());
final String formatStr = ".Format: '";
final String desc = infoHeader.getDescription();
int i = desc.indexOf(formatStr);
if (i != -1) i = desc.indexOf('(', i + formatStr.length());
int j = desc.lastIndexOf(')');
if (i != -1 && j > i) {
snpEffColumns =
pipe.split(desc.substring(i + 1, j).replaceAll("[ \\[\\]()\\.]", "").trim());
LOG.info(Arrays.asList(snpEffColumns));
} else {
LOG.error("Cannot parse " + infoHeader.getDescription());
}
}
String nmdColumns[] = null;
infoHeader = header.getInfoHeaderLine("NMD");
if (infoHeader != null && this.USE_SNPEFF) {
final String formatStr = " Format: '";
final String desc = infoHeader.getDescription();
int i = desc.indexOf(formatStr);
int j = (i == -1 ? -1 : desc.lastIndexOf('\''));
if (i != -1 && j > i) {
nmdColumns =
pipe.split(
desc.substring(i + formatStr.length(), j).replaceAll("[ \\[\\]()\\.]", "").trim());
} else {
LOG.error("Cannot parse " + infoHeader.getDescription());
}
}
String lofColumns[] = null;
infoHeader = header.getInfoHeaderLine("LOF");
if (infoHeader != null && this.USE_SNPEFF) {
final String formatStr = " Format: '";
final String desc = infoHeader.getDescription();
int i = desc.indexOf(formatStr);
int j = (i == -1 ? -1 : desc.lastIndexOf('\''));
if (i != -1 && j > i) {
lofColumns =
pipe.split(
desc.substring(i + formatStr.length(), j).replaceAll("[ \\[\\]()\\.]", "").trim());
} else {
LOG.error("Cannot parse " + infoHeader.getDescription());
}
}
for (String S : header.getSampleNamesInOrder()) {
// merge into SAMPLE using (select 1+MAX(id),'azdazd' from SAMPLE) as vals(x,y) on
// SAMPLE.name=vals.y when NOT MATCHED THEN INSERT VALUES vals.x,vals.y;
switch (this.engine) {
case hsql:
out.println(
"merge into SAMPLE"
+ SUFFIX
+ " using ( values("
+ quote(S)
+ ") ) "
+ "AS vals(y) ON SAMPLE"
+ SUFFIX
+ ".name = vals.y "
+ "WHEN NOT MATCHED THEN INSERT VALUES (NULL,vals.y);");
break;
default:
out.println(
"insert or ignore into SAMPLE" + SUFFIX + "(name) values (" + quote(S) + ");");
break;
}
}
List<String> headers = new ArrayList<String>();
for (VCFHeaderLine line : header.getMetaDataInSortedOrder()) {
if (VCFHeaderVersion.isFormatString(line.getKey())) continue;
headers.add(VCFHeader.METADATA_INDICATOR + line);
}
String chromLine = VCFHeader.HEADER_INDICATOR;
for (VCFHeader.HEADER_FIELDS field : header.getHeaderFields()) {
if (!VCFHeader.HEADER_INDICATOR.equals(chromLine))
chromLine += (VCFConstants.FIELD_SEPARATOR);
chromLine += (field);
}
if (header.hasGenotypingData()) {
chromLine += VCFConstants.FIELD_SEPARATOR + "FORMAT";
for (String sample : header.getGenotypeSamples()) {
chromLine += VCFConstants.FIELD_SEPARATOR;
chromLine += sample;
}
}
headers.add(chromLine);
for (String line : headers) {
out.println(
"insert into HEADER"
+ SUFFIX
+ "(file_id,header) values ("
+ "(select max(id) from FILE"
+ SUFFIX
+ "),"
+ quote(line)
+ ");");
}
while (r.hasNext()) {
VariantContext var = r.next();
if (var == null) {
LOG.error("Cannot parse VCF");
continue;
}
// "create table if not exists FILE(id,filename text)";
// "create table if not exists VARIATION(id,file_id,chrom,pos,start0,end0,rs_id,ref,qual)";
out.println(
"insert into VARIATION"
+ SUFFIX
+ "(file_id,chrom,pos,START0,END0,rs_id,ref,qual) values ("
+ "(select max(id) from FILE"
+ SUFFIX
+ "),"
+ quote(var.getChr())
+ ","
+ var.getStart()
+ ","
+ (var.getStart() - 1)
+ ","
+ var.getEnd()
+ ","
+ (var.getID() == null || var.getID().equals(VCFConstants.EMPTY_ID_FIELD)
? "NULL"
: quote(var.getID()))
+ ","
+ quote(var.getReference().getDisplayString())
+ ","
+ (var.getPhredScaledQual() < 0 ? "NULL" : var.getPhredScaledQual())
+ ");");
// "create table if not exists ALT(id,var_id,alt)";
for (Allele alt : var.getAlternateAlleles()) {
out.println(
"insert into ALT"
+ SUFFIX
+ "(var_id,alt) values ("
+ "(select max(id) from VARIATION"
+ SUFFIX
+ "),"
+ quote(alt.getDisplayString())
+ ");");
}
// "create table if not exists FILTER(id,var_id,filter)";
for (String filter : var.getFilters()) {
out.println(
"insert into FILTER"
+ SUFFIX
+ "(var_id,filter) values ("
+ "(select max(id) from VARIATION"
+ SUFFIX
+ "),"
+ quote(filter)
+ ");");
}
CommonInfo infos = var.getCommonInfo();
for (String key : infos.getAttributes().keySet()) {
Object val = infos.getAttribute(key);
// "create table if not exists INFO(id,var_id,k,v)";
if (SPLIT4 && key.equals("DP4")) {
String dp4[] = infotoString(val).split("[,]");
insertIntoInfo(quote(key + "[refFor]"), quote(dp4[0]));
insertIntoInfo(quote(key + "[refRev]"), quote(dp4[1]));
insertIntoInfo(quote(key + "[altFor]"), quote(dp4[2]));
insertIntoInfo(quote(key + "[altRev]"), quote(dp4[3]));
} else {
insertIntoInfo(quote(key), quote(infotoString(val)));
}
if (key.equals("CSQ") && csqColumns != null) {
List as_array = castToStringArray(val);
for (Object csqs : as_array) {
if (csqs.toString().isEmpty()) continue;
String tokens[] = pipe.split(csqs.toString());
List<String> extraInfo = new ArrayList<String>();
for (int t = 0; t < tokens.length && t < csqColumns.length; ++t) {
if (tokens[t].isEmpty()) continue;
if (csqColumns[t].equals("Consequence")) {
for (String pred : amp.split(tokens[t])) {
if (pred.isEmpty()) continue;
extraInfo.add(csqColumns[t]);
extraInfo.add(pred);
}
} else {
extraInfo.add(csqColumns[t]);
extraInfo.add(tokens[t]);
}
}
insertExtraInfos("CSQ", extraInfo);
}
}
if (key.equals("EFF") && snpEffColumns != null) {
for (Object item : castToStringArray(val)) {
String snpeff = item.toString();
if (snpeff.isEmpty()) continue;
int opar = snpeff.indexOf('(');
if (opar == -1) continue;
int cpar = snpeff.lastIndexOf(')');
if (cpar == -1) continue;
String tokens[] = pipe.split(snpeff.substring(opar + 1, cpar));
List<String> h = new ArrayList<String>();
h.add("Effect");
h.add(snpeff.substring(0, opar));
for (int t = 0; t < tokens.length && t < snpEffColumns.length; ++t) {
if (tokens[t].isEmpty()) continue;
h.add(snpEffColumns[t]);
h.add(tokens[t]);
}
insertExtraInfos(key, h);
}
}
if (key.equals("NMD") && nmdColumns != null) {
for (Object item : castToStringArray(val)) {
String nmd = item.toString();
if (nmd.isEmpty()) continue;
String tokens[] = pipe.split(nmd);
List<String> h = new ArrayList<String>(nmdColumns.length * 2);
for (int t = 0; t < tokens.length && t < nmdColumns.length; ++t) {
if (tokens[t].isEmpty()) continue;
h.add(nmdColumns[t]);
h.add(tokens[t]);
}
insertExtraInfos(key, h);
}
}
if (key.equals("LOF") && lofColumns != null) {
for (Object item : castToStringArray(val)) {
String lof = item.toString();
if (lof.isEmpty()) continue;
String tokens[] = pipe.split(lof);
List<String> h = new ArrayList<String>(lofColumns.length * 2);
for (int t = 0; t < tokens.length && t < lofColumns.length; ++t) {
if (tokens[t].isEmpty()) continue;
h.add(lofColumns[t]);
h.add(tokens[t]);
}
insertExtraInfos(key, h);
}
}
}
GenotypesContext genotypesCtx = var.getGenotypes();
for (Genotype g : genotypesCtx) {
// "create table if not exists GENOTYPE(id,var_id,k,v)";
List<Allele> alleles = g.getAlleles();
out.println(
"insert into GENOTYPE"
+ SUFFIX
+ "(var_id,sample_id,A1,A2,dp,ad,gq,pl,"
+ "is_phased,is_hom,is_homref,is_homvar,is_mixed,"
+ "is_nocall,is_noninformative,is_available,is_called,is_filtered"
+ ") values ("
+ "(select max(id) from VARIATION"
+ SUFFIX
+ "),"
+ "(select id from SAMPLE"
+ SUFFIX
+ " where name="
+ quote(g.getSampleName())
+ "),"
+ (alleles.size() == 2 ? quote(alleles.get(0).getBaseString()) : "NULL")
+ ","
+ (alleles.size() == 2 ? quote(alleles.get(1).getBaseString()) : "NULL")
+ ","
+ (g.hasDP() ? g.getDP() : "NULL")
+ ","
+ (g.hasAD() ? quote(infotoString(g.getAD())) : "NULL")
+ ","
+ (g.hasGQ() ? g.getGQ() : "NULL")
+ ","
+ (g.hasPL() ? quote(infotoString(g.getPL())) : "NULL")
+ ","
+ (g.isPhased() ? 1 : 0)
+ ","
+ (g.isHom() ? 1 : 0)
+ ","
+ (g.isHomRef() ? 1 : 0)
+ ","
+ (g.isHomVar() ? 1 : 0)
+ ","
+ (g.isMixed() ? 1 : 0)
+ ","
+ (g.isNoCall() ? 1 : 0)
+ ","
+ (g.isNonInformative() ? 1 : 0)
+ ","
+ (g.isAvailable() ? 1 : 0)
+ ","
+ (g.isCalled() ? 1 : 0)
+ ","
+ (g.isFiltered() ? 1 : 0)
+ ");");
for (String key : g.getExtendedAttributes().keySet()) {
Object val = g.getExtendedAttribute(key);
if (val == null) continue;
out.println(
"insert into GTPROP"
+ SUFFIX
+ "(genotype_id,k,v) values ("
+ "(select max(id) from GENOTYPE"
+ SUFFIX
+ "),"
+ quote(key)
+ ","
+ quote(infotoString(val))
+ ");");
}
}
}
r.close();
}
