From a4ae5d554a09ef724be2702b876ff6da571529b8 Mon Sep 17 00:00:00 2001
From: Olivier Langella <olivier.langella@u-psud.fr>
Date: Mon, 5 Mar 2018 22:11:23 +0100
Subject: [PATCH] WIP: Mascot parser
---
src/input/mascot/mascotdatparser.cpp | 248 ++++++++++++++++++++-------
src/input/mascot/mascotdatparser.h | 17 ++
src/utils/types.h | 3 +-
3 files changed, 203 insertions(+), 65 deletions(-)
diff --git a/src/input/mascot/mascotdatparser.cpp b/src/input/mascot/mascotdatparser.cpp
index 1ef92308c..fc06c24f2 100644
--- a/src/input/mascot/mascotdatparser.cpp
+++ b/src/input/mascot/mascotdatparser.cpp
@@ -33,6 +33,7 @@
#include <pappsomspp/pappsoexception.h>
#include "../../core/peptidextp.h"
#include "../../core/proteinmatch.h"
+#include "../../core/peptideevidence.h"
MascotDatParser::MascotDatParser(Project * p_project, IdentificationGroup * p_identification_group,
IdentificationDataSource * p_identification_data_source) {
@@ -63,11 +64,23 @@ void MascotDatParser::parse(QIODevice * in_stream) {
parseHeaderLine( mime_parser.getCurrentTextStream().readLine());
}
}
+ else if (mime_parser.getCurrentFileName() == "summary") {
+ while(!mime_parser.getCurrentTextStream().atEnd()) {
+ parseSummaryLine( mime_parser.getCurrentTextStream().readLine());
+ }
+ }
else if (mime_parser.getCurrentFileName() == "peptides") {
while(!mime_parser.getCurrentTextStream().atEnd()) {
parsePeptidesLine( mime_parser.getCurrentTextStream().readLine());
}
}
+ else if (mime_parser.getCurrentFileName().startsWith("query")) {
+ _current_query_index = mime_parser.getCurrentFileName().mid(5).toUInt();
+ while(!mime_parser.getCurrentTextStream().atEnd()) {
+ parseQueryLine( mime_parser.getCurrentTextStream().readLine());
+ }
+ saveQuery();
+ }
}
mime_parser.close();
@@ -122,6 +135,7 @@ void MascotDatParser::parseHeaderLine(const QString & header_line) {
qDebug() << "queries=" << header_list[2];
_number_of_queries = header_list[2].toUInt();
_query_peptide_results.resize(_number_of_queries);
+ _summary_list.resize(_number_of_queries);
}
//min_peaks_for_homology=6
//max_hits=50
@@ -149,70 +163,6 @@ void MascotDatParser::parseHeaderLine(const QString & header_line) {
}
}
-void MascotDatParser::saveAndClearPeptide() {
- qDebug() << "MascotDatParser::saveAndClearPeptide begin";
- if (_current_peptide.query_index > 0) {
- // save
- _query_peptide_results[_current_peptide.query_index-1].push_back(_current_peptide);
- //parse and save
-
- QString peptide_str =_current_peptide.peptide_string_list.at(4);
- if (!_current_peptide.subst.isEmpty()) {
- //q856_p9_subst=1,X,W
- //q24379_p2_subst=1,B,D,8,B,D
- QStringList subst_list = _current_peptide.subst.split(",");
- for (unsigned int i=0; i < subst_list.size(); i+=3) {
- peptide_str = peptide_str.replace(subst_list.at(0+i).toInt()-1,1,subst_list.at(2+i));
- }
- }
- PeptideXtpSp peptide_sp;
- peptide_sp = PeptideXtp(peptide_str).makePeptideXtpSp();
- peptide_sp = _p_project->getPeptideStore().getInstance(peptide_sp);
-
- if (_current_peptide.protein_string_list.size() != _current_peptide.fasta_file_list.size()) {
- throw pappso::PappsoException(QObject::tr("ERROR (_current_peptide.protein_string_list.size() != _current_peptide.fasta_file_list.size()) %1").arg(_current_peptide.protein_string_list.join(",\"")));
- }
-
- foreach (const QString &str, _current_peptide.protein_string_list) {
- //sp|O95006|OR2F2_HUMAN":0:299:303:1
- int position = str.indexOf("\"", 0);
- QString accession = str.mid(0, position);
- qDebug() << "accession=" << accession;
- QStringList position_list = str.mid(position+2).split(":");
- if (position_list.size() != 4) {
- throw pappso::PappsoException(QObject::tr("ERROR position_list.size() != 4 %1").arg(str));
- }
- unsigned int start = position_list.at(1).toUInt();
- unsigned int stop = position_list.at(2).toUInt();
-
- ProteinXtp protein;
- protein.setAccession(accession);
-
- ProteinMatch * p_protein_match = _p_identification_group->getProteinMatchInstance(protein.getAccession());
- if (p_protein_match == nullptr) {
- throw pappso::PappsoException(QObject::tr("ERROR (p_protein_match == nullptr) %1").arg(str));
- }
-
- ProteinXtpSp sp_xtp_protein = protein.makeProteinXtpSp();
- p_protein_match->setProteinXtpSp(_p_project->getProteinStore().getInstance(sp_xtp_protein));
- p_protein_match->setChecked(true);
-
- PeptideMatch peptide_match;
- peptide_match.setStart(start);
- //peptide_match.setPeptideEvidenceSp();
-
- p_protein_match->addPeptideMatch(peptide_match);
- }
- }
-
- //new peptide query clear
- _current_peptide.peptide_string_list.clear();
- _current_peptide.fasta_file_list.clear();
- _current_peptide.query_index = 0;
- _current_peptide.subst = "";
- qDebug() << "MascotDatParser::saveAndClearPeptide end";
-}
-
void MascotDatParser::parsePeptidesLine(const QString & peptide_line) {
try {
if (_regexp_header_line.exactMatch(peptide_line)) {
@@ -304,3 +254,173 @@ void MascotDatParser::parsePeptidesLine(const QString & peptide_line) {
throw pappso::PappsoException(_error_str);
}
}
+
+
+void MascotDatParser::parseQueryLine(const QString & query_line) {
+ try {
+ if (_regexp_header_line.exactMatch(query_line)) {
+ QStringList header_list = _regexp_header_line.capturedTexts();
+ QString index = header_list[1];
+ QString value = header_list[2];
+
+
+ //title=FULL%20ISSLGSVGAGIVAVKK%20N22213%20%20%20QEP1_SpikeIn_230914_1_3ng_270914%2e35282%2e35282%2e2
+ if (index == "title") {
+ _current_query.title = value;
+ }
+ //rtinseconds=5703.84
+ else if (index == "rtinseconds") {
+ _current_query.rt = value.toDouble();
+ }
+ //index=44035
+ else if (index == "index") {
+ _current_query.query_index = value.toUInt();
+ }
+ //charge=2+
+ else if (index == "charge") {
+ _current_query.charge = value.toUInt();
+ }
+ }
+ /*
+ mass_min=129.102051
+ mass_max=1198.751099
+ int_min=2327
+ int_max=6.845e+005
+ num_vals=44
+ num_used1=-1
+ Ions1=129.102051:1.111e+005,275.207306:9.008e+004,374.275299:9.717e+004,514.825623:1.929e+004,599.878906:1.472e+005,714.431519:1.356e+004,815.466003:9643,924.570435:4428,1028.645630:1.257e+005,1085.665527:6.845e+005,1198.751099:1.858e+005,147.112656:7.954e+004,261.159271:4.327e+004,357.248169:2.192e+004,506.312775:3319,543.336548:1.083e+005,655.376465:9786,1011.625977:1.747e+004,1068.641602:9.804e+004,1181.725342:2.712e+004,130.086014:3.602e+004,257.197388:1.455e+004,357.212097:3960,591.370056:2.454e+004,972.547058:3819,1069.647583:4.457e+004,201.123077:1.566e+004,299.171783:1.421e+004,534.825562:1.036e+004,1087.671631:2.374e+004,173.128418:1.512e+004,244.129089:1.052e+004,590.873291:8065,1067.656006:2.235e+004,228.133957:1.35e+004,258.180115:3194,534.333374:3635,1096.635376:1.549e+004,200.139084:1.026e+004,535.328552:3006,1070.637085:4813,183.112442:9925,131.080994:4806,211.108139:2327
+
+ */
+ }
+
+ catch (pappso::PappsoException exception_pappso) {
+ _error_str = QObject::tr("ERROR in MascotDatParser::parseQueryLine %1, PAPPSO exception:\n%2").arg(query_line).arg(exception_pappso.qwhat());
+ qDebug() << _error_str;
+ throw pappso::PappsoException(_error_str);
+ }
+ catch (std::exception exception_std) {
+ _error_str = QObject::tr("ERROR in MascotDatParser::parseQueryLine %1, std exception:\n%2").arg(query_line).arg(exception_std.what());
+ qDebug() << _error_str;
+ throw pappso::PappsoException(_error_str);
+ }
+}
+
+
+void MascotDatParser::parseSummaryLine(const QString & summary_line) {
+
+ if (_regexp_header_line.exactMatch(summary_line)) {
+ QStringList header_list = _regexp_header_line.capturedTexts();
+ QString index = header_list[1];
+ QString value = header_list[2];
+
+
+ //qmass1=598.300206
+ if (index.startsWith("qmass")) {
+ unsigned int query_index = index.mid(5).toUInt();
+ //_current_query.title = value;
+ _summary_list[query_index-1].exp_mass=value.toDouble();
+ }
+ //qexp1=300.157379,2+
+//qintensity1=2054822.6250
+//qmatch1=73
+ else if (index.startsWith("qmatch")) {
+ unsigned int query_index = index.mid(6).toUInt();
+ //_current_query.title = value;
+ _summary_list[query_index-1].match=value.toUInt();
+ }
+//qplughole1=14.820890
+ else if (index.startsWith("qplughole")) {
+ unsigned int query_index = index.mid(9).toUInt();
+ //_current_query.title = value;
+ _summary_list[query_index-1].plug_hole=value.toDouble();
+ }
+
+ }
+}
+
+
+void MascotDatParser::saveAndClearPeptide() {
+ qDebug() << "MascotDatParser::saveAndClearPeptide begin";
+ if (_current_peptide.query_index > 0) {
+ // save
+ _query_peptide_results[_current_peptide.query_index-1].push_back(_current_peptide);
+ }
+
+ //new peptide query clear
+ _current_peptide.peptide_string_list.clear();
+ _current_peptide.fasta_file_list.clear();
+ _current_peptide.query_index = 0;
+ _current_peptide.subst = "";
+ qDebug() << "MascotDatParser::saveAndClearPeptide end";
+}
+
+
+void MascotDatParser::saveQuery() {
+ qDebug() << "MascotDatParser::saveQuery begin";
+ if (_current_query.query_index > 0) {
+ std::vector< PeptideLine> peptide_list = _query_peptide_results[_current_query.query_index-1];
+
+ PeptideEvidence peptide_evidence(_p_identification_data_source->getMsRunSp().get(),_current_query.query_index);
+ peptide_evidence.setCharge(_current_query.charge);
+ peptide_evidence.setChecked(true);
+ peptide_evidence.setExperimentalMass(_summary_list[_current_query.query_index-1].exp_mass);
+ peptide_evidence.setRetentionTime(_current_query.rt);
+
+ peptide_evidence.setParam(PeptideEvidenceParam::mascot_macth_score, QVariant(_summary_list[_current_query.query_index-1].match));
+
+ peptide_evidence.setIdentificationDataSource( _p_identification_data_source);
+ //parse and save
+ for(PeptideLine & peptide_line:peptide_list) {
+
+ QString peptide_str =peptide_line.peptide_string_list.at(4);
+ if (!peptide_line.subst.isEmpty()) {
+ //q856_p9_subst=1,X,W
+ //q24379_p2_subst=1,B,D,8,B,D
+ QStringList subst_list = peptide_line.subst.split(",");
+ for (unsigned int i=0; i < subst_list.size(); i+=3) {
+ peptide_str = peptide_str.replace(subst_list.at(0+i).toInt()-1,1,subst_list.at(2+i));
+ }
+ }
+ PeptideXtpSp peptide_sp;
+ peptide_sp = PeptideXtp(peptide_str).makePeptideXtpSp();
+ peptide_sp = _p_project->getPeptideStore().getInstance(peptide_sp);
+
+ if (peptide_line.protein_string_list.size() != peptide_line.fasta_file_list.size()) {
+ throw pappso::PappsoException(QObject::tr("ERROR (peptide_line.protein_string_list.size() != peptide_line.fasta_file_list.size()) %1").arg(peptide_line.protein_string_list.join(",\"")));
+ }
+
+ foreach (const QString &str, peptide_line.protein_string_list) {
+ //sp|O95006|OR2F2_HUMAN":0:299:303:1
+ int position = str.indexOf("\"", 0);
+ QString accession = str.mid(0, position);
+ qDebug() << "accession=" << accession;
+ QStringList position_list = str.mid(position+2).split(":");
+ if (position_list.size() != 4) {
+ throw pappso::PappsoException(QObject::tr("ERROR position_list.size() != 4 %1").arg(str));
+ }
+ unsigned int start = position_list.at(1).toUInt();
+ unsigned int stop = position_list.at(2).toUInt();
+
+ ProteinXtp protein;
+ protein.setAccession(accession);
+
+ ProteinMatch * p_protein_match = _p_identification_group->getProteinMatchInstance(protein.getAccession());
+ if (p_protein_match == nullptr) {
+ throw pappso::PappsoException(QObject::tr("ERROR (p_protein_match == nullptr) %1").arg(str));
+ }
+
+ ProteinXtpSp sp_xtp_protein = protein.makeProteinXtpSp();
+ p_protein_match->setProteinXtpSp(_p_project->getProteinStore().getInstance(sp_xtp_protein));
+ p_protein_match->setChecked(true);
+
+ PeptideMatch peptide_match;
+ peptide_match.setStart(start);
+ peptide_match.setPeptideEvidenceSp(_p_identification_data_source->getPeptideEvidenceStore().getInstance(&peptide_evidence));
+
+ p_protein_match->addPeptideMatch(peptide_match);
+ }
+ }
+ }
+
+ qDebug() << "MascotDatParser::saveQuery end";
+}
diff --git a/src/input/mascot/mascotdatparser.h b/src/input/mascot/mascotdatparser.h
index 23f387754..d4ef5fd69 100644
--- a/src/input/mascot/mascotdatparser.h
+++ b/src/input/mascot/mascotdatparser.h
@@ -43,6 +43,9 @@ private:
void parseProteinLine(const QString & protein_line);
void parseHeaderLine(const QString & protein_line);
void parsePeptidesLine(const QString & peptide_line);
+ void parseQueryLine(const QString & query_line);
+ void parseSummaryLine(const QString & summary_line);
+ void saveQuery();
void saveAndClearPeptide();
struct PeptideLine {
@@ -53,6 +56,17 @@ private:
QStringList protein_string_list;
std::vector<FastaFileSp> fasta_file_list;
};
+ struct QueryLine {
+ unsigned int query_index=0;
+ unsigned int charge=0;
+ pappso::pappso_double rt=0;
+ QString title;
+ };
+ struct SummaryLine {
+ unsigned int match=0;
+ pappso::pappso_double exp_mass=0;
+ pappso::pappso_double plug_hole=0;
+ };
private:
Project * _p_project;
IdentificationGroup * _p_identification_group;
@@ -67,11 +81,14 @@ private:
QRegExp _regexp_header_line;
unsigned int _number_of_queries=0;
unsigned int _number_of_residues=0;
+ unsigned int _current_query_index=0;
QString _error_str;
PeptideLine _current_peptide;
+ QueryLine _current_query;
std::vector<std::vector<PeptideLine>> _query_peptide_results;
+ std::vector<SummaryLine> _summary_list;
};
diff --git a/src/utils/types.h b/src/utils/types.h
index c228b07d8..26bc2fd44 100644
--- a/src/utils/types.h
+++ b/src/utils/types.h
@@ -54,7 +54,8 @@ enum class IdentificationEngine: std::int8_t {
*
*/
enum class PeptideEvidenceParam: std::int8_t {
- tandem_hyperscore=0 ///< X!Tandem hyperscore
+ tandem_hyperscore=0, ///< X!Tandem hyperscore
+ mascot_macth_score=1 ///< MASCOT match score
};
/** \def IdentificationEngineParam identification engine parameters
--
GitLab