1 // ***************************************************************************
2 // BamMultiReader.cpp (c) 2010 Erik Garrison
3 // Marth Lab, Department of Biology, Boston College
4 // All rights reserved.
5 // ---------------------------------------------------------------------------
6 // Last modified: 23 Februrary 2010 (EG)
7 // ---------------------------------------------------------------------------
8 // Uses BGZF routines were adapted from the bgzf.c code developed at the Broad
10 // ---------------------------------------------------------------------------
11 // Functionality for simultaneously reading multiple BAM files.
13 // This functionality allows applications to work on very large sets of files
14 // without requiring intermediate merge, sort, and index steps for each file
15 // subset. It also improves the performance of our merge system as it
16 // precludes the need to sort merged files.
17 // ***************************************************************************
29 #include "BamMultiReader.h"
30 using namespace BamTools;
33 // -----------------------------------------------------
34 // BamMultiReader implementation
35 // -----------------------------------------------------
38 BamMultiReader::BamMultiReader(void)
44 BamMultiReader::~BamMultiReader(void) {
45 Close(); // close the bam files
46 // clean up reader objects
47 for (vector<pair<BamReader*, BamAlignment*> >::iterator it = readers.begin(); it != readers.end(); ++it) {
53 // close the BAM files
54 void BamMultiReader::Close(void) {
55 for (vector<pair<BamReader*, BamAlignment*> >::iterator it = readers.begin(); it != readers.end(); ++it) {
56 BamReader* reader = it->first;
57 reader->Close(); // close the reader
61 // updates the reference id stored in the BamMultiReader
62 // to reflect the current state of the readers
63 void BamMultiReader::UpdateReferenceID(void) {
64 // the alignments are sorted by position, so the first alignment will always have the lowest reference ID
65 if (alignments.begin()->second.second->RefID != CurrentRefID) {
66 // get the next reference id
67 // while there aren't any readers at the next ref id
68 // increment the ref id
69 int nextRefID = CurrentRefID;
70 while (alignments.begin()->second.second->RefID != nextRefID) {
73 //cerr << "updating reference id from " << CurrentRefID << " to " << nextRefID << endl;
74 CurrentRefID = nextRefID;
78 // checks if any readers still have alignments
79 bool BamMultiReader::HasOpenReaders() {
80 return alignments.size() > 0;
83 // get next alignment among all files (from specified region, if given)
84 bool BamMultiReader::GetNextAlignment(BamAlignment& nextAlignment) {
86 // bail out if we are at EOF in all files, means no more alignments to process
87 if (!HasOpenReaders())
90 // when all alignments have stepped into a new target sequence, update our
91 // current reference sequence id
94 // our lowest alignment and reader will be at the front of our alignment index
95 BamAlignment* lowestAlignment = alignments.begin()->second.second;
96 BamReader* lowestReader = alignments.begin()->second.first;
98 // now that we have the lowest alignment in the set, save it by copy to our argument
99 nextAlignment = BamAlignment(*lowestAlignment);
101 // remove this alignment index entry from our alignment index
102 alignments.erase(alignments.begin());
104 // and add another entry if we can get another alignment from the reader
105 if (lowestReader->GetNextAlignment(*lowestAlignment)) {
106 alignments.insert(make_pair(make_pair(lowestAlignment->RefID, lowestAlignment->Position),
107 make_pair(lowestReader, lowestAlignment)));
108 } else { // do nothing
109 //cerr << "reached end of file " << lowestReader->GetFilename() << endl;
115 // jumps to specified region(refID, leftBound) in BAM files, returns success/fail
116 bool BamMultiReader::Jump(int refID, int position) {
118 //if ( References.at(refID).RefHasAlignments && (position <= References.at(refID).RefLength) ) {
119 CurrentRefID = refID;
120 CurrentLeft = position;
123 for (vector<pair<BamReader*, BamAlignment*> >::iterator it = readers.begin(); it != readers.end(); ++it) {
124 BamReader* reader = it->first;
125 result &= reader->Jump(refID, position);
127 cerr << "ERROR: could not jump " << reader->GetFilename() << " to " << refID << ":" << position << endl;
134 for (vector<pair<BamReader*, BamAlignment*> >::iterator it = readers.begin(); it != readers.end(); ++it) {
135 BamReader* br = it->first;
136 BamAlignment* ba = it->second;
137 if (br->GetNextAlignment(*ba)) {
138 alignments.insert(make_pair(make_pair(ba->RefID, ba->Position),
141 // assume BamReader EOF
149 void BamMultiReader::Open(const vector<string> filenames, bool openIndexes) {
150 // for filename in filenames
151 fileNames = filenames; // save filenames in our multireader
152 for (vector<string>::const_iterator it = filenames.begin(); it != filenames.end(); ++it) {
153 string filename = *it;
154 BamReader* reader = new BamReader;
156 reader->Open(filename, filename + ".bai");
158 reader->Open(filename); // for merging, jumping is disallowed
160 BamAlignment* alignment = new BamAlignment;
161 reader->GetNextAlignment(*alignment);
162 readers.push_back(make_pair(reader, alignment)); // store pointers to our readers for cleanup
163 alignments.insert(make_pair(make_pair(alignment->RefID, alignment->Position),
164 make_pair(reader, alignment)));
169 void BamMultiReader::PrintFilenames(void) {
170 for (vector<pair<BamReader*, BamAlignment*> >::iterator it = readers.begin(); it != readers.end(); ++it) {
171 BamReader* reader = it->first;
172 cout << reader->GetFilename() << endl;
177 void BamMultiReader::DumpAlignmentIndex(void) {
178 for (AlignmentIndex::const_iterator it = alignments.begin(); it != alignments.end(); ++it) {
179 cerr << it->first.first << ":" << it->first.second << " " << it->second.first->GetFilename() << endl;
183 // returns BAM file pointers to beginning of alignment data
184 bool BamMultiReader::Rewind(void) {
186 for (vector<pair<BamReader*, BamAlignment*> >::iterator it = readers.begin(); it != readers.end(); ++it) {
187 BamReader* reader = it->first;
188 result &= reader->Rewind();
193 // saves index data to BAM index files (".bai") where necessary, returns success/fail
194 bool BamMultiReader::CreateIndexes(void) {
196 for (vector<pair<BamReader*, BamAlignment*> >::iterator it = readers.begin(); it != readers.end(); ++it) {
197 BamReader* reader = it->first;
198 result &= reader->CreateIndex();
203 // makes a virtual, unified header for all the bam files in the multireader
204 const string BamMultiReader::GetHeaderText(void) const {
206 string mergedHeader = "";
207 map<string, bool> readGroups;
209 // foreach extraction entry (each BAM file)
210 bool isFirstTime = true;
211 for (vector<pair<BamReader*, BamAlignment*> >::const_iterator it = readers.begin(); it != readers.end(); ++it) {
213 BamReader* reader = it->first;
215 stringstream header(reader->GetHeaderText());
216 vector<string> lines;
218 while (getline(header, item))
219 lines.push_back(item);
221 for (vector<string>::const_iterator it = lines.begin(); it != lines.end(); ++it) {
223 // get next line from header, skip if empty
224 string headerLine = *it;
225 if ( headerLine.empty() ) { continue; }
227 // if first file, save HD & SQ entries
229 if ( headerLine.find("@HD") == 0 || headerLine.find("@SQ") == 0) {
230 mergedHeader.append(headerLine.c_str());
231 mergedHeader.append(1, '\n');
235 // (for all files) append RG entries if they are unique
236 if ( headerLine.find("@RG") == 0 ) {
237 stringstream headerLineSs(headerLine);
238 string part, readGroupPart, readGroup;
239 while(std::getline(headerLineSs, part, '\t')) {
241 std::getline(headerLineSs, readGroupPart, '\t');
242 stringstream readGroupPartSs(readGroupPart);
243 std::getline(readGroupPartSs, readGroup, ':');
244 std::getline(readGroupPartSs, readGroup, ':');
248 if (readGroups.find(readGroup) == readGroups.end()) { // prevents duplicate @RG entries
249 mergedHeader.append(headerLine.c_str() );
250 mergedHeader.append(1, '\n');
251 readGroups[readGroup] = true;
257 // set iteration flag
261 // return merged header text
265 // ValidateReaders checks that all the readers point to BAM files representing
266 // alignments against the same set of reference sequences, and that the
267 // sequences are identically ordered. If these checks fail the operation of
268 // the multireader is undefined, so we force program exit.
269 void BamMultiReader::ValidateReaders(void) const {
270 int firstRefCount = readers.front().first->GetReferenceCount();
271 BamTools::RefVector firstRefData = readers.front().first->GetReferenceData();
272 for (vector<pair<BamReader*, BamAlignment*> >::const_iterator it = readers.begin(); it != readers.end(); ++it) {
273 BamReader* reader = it->first;
274 BamTools::RefVector currentRefData = reader->GetReferenceData();
275 BamTools::RefVector::const_iterator f = firstRefData.begin();
276 BamTools::RefVector::const_iterator c = currentRefData.begin();
277 if (reader->GetReferenceCount() != firstRefCount || firstRefData.size() != currentRefData.size()) {
278 cerr << "ERROR: mismatched number of references in " << reader->GetFilename()
279 << " expected " << firstRefCount
280 << " reference sequences but only found " << reader->GetReferenceCount() << endl;
283 // this will be ok; we just checked above that we have identically-sized sets of references
284 // here we simply check if they are all, in fact, equal in content
285 while (f != firstRefData.end()) {
286 if (f->RefName != c->RefName || f->RefLength != c->RefLength) {
287 cerr << "ERROR: mismatched references found in " << reader->GetFilename()
288 << " expected: " << endl;
289 for (BamTools::RefVector::const_iterator a = firstRefData.begin(); a != firstRefData.end(); ++a)
290 cerr << a->RefName << " " << a->RefLength << endl;
291 cerr << "but found: " << endl;
292 for (BamTools::RefVector::const_iterator a = currentRefData.begin(); a != currentRefData.end(); ++a)
293 cerr << a->RefName << " " << a->RefLength << endl;
301 // NB: The following functions assume that we have identical references for all
302 // BAM files. We enforce this by invoking the above validation function
303 // (ValidateReaders) to verify that our reference data is the same across all
304 // files on Open, so we will not encounter a situation in which there is a
305 // mismatch and we are still live.
307 // returns the number of reference sequences
308 const int BamMultiReader::GetReferenceCount(void) const {
309 return readers.front().first->GetReferenceCount();
312 // returns vector of reference objects
313 const BamTools::RefVector BamMultiReader::GetReferenceData(void) const {
314 return readers.front().first->GetReferenceData();
317 const int BamMultiReader::GetReferenceID(const string& refName) const {
318 return readers.front().first->GetReferenceID(refName);