TranscriptGeneMap transcriptToGeneMapFromFasta( const std::string& transcriptsFile ) {
using std::vector;
NameVector transcriptNames;
NameVector geneNames {"gene"};
vector<bfs::path> paths{transcriptsFile};
StreamingReadParser parser(paths);
parser.start();
ReadProducer<StreamingReadParser> producer(parser);
ReadSeq* s;
// while there are transcripts left to process
while (producer.nextRead(s)) {
// The transcript name
std::string fullHeader(s->name, s->nlen);
std::string header = fullHeader.substr(0, fullHeader.find(' '));
transcriptNames.emplace_back(header);
producer.finishedWithRead(s);
}
// Sort the transcript names
std::sort(transcriptNames.begin(), transcriptNames.end());
// Since we have no real gene groupings, the t2g vector is trivial,
// everything maps to gene 0.
IndexVector t2g(transcriptNames.size(), 0);
return TranscriptGeneMap(transcriptNames, geneNames, t2g);
}
void _computeReverseMap() {
_genesToTranscripts.resize( _geneNames.size(), {});
Index geneID;
Index transcriptID = 0;
size_t maxNumTrans = 0;
Index maxGene;
for ( size_t transcriptID = 0; transcriptID < _transcriptsToGenes.size(); ++transcriptID ) {
_genesToTranscripts[ _transcriptsToGenes[transcriptID] ].push_back( transcriptID );
if ( maxNumTrans < _genesToTranscripts[ _transcriptsToGenes[transcriptID] ].size() ) {
maxNumTrans = _genesToTranscripts[ _transcriptsToGenes[transcriptID] ].size();
maxGene = _transcriptsToGenes[transcriptID];
}
}
std::cerr << "max # of transcripts in a gene was " << maxNumTrans << " in gene " << _geneNames[maxGene] << "\n";
}
Size numGenes() {
return _geneNames.size();
}
Size numTranscripts() {
return _transcriptNames.size();
}
