IDScoreSwitcherAlgorithm.h

Go to the documentation of this file.

// Copyright (c) 2002-present, OpenMS Inc. -- EKU Tuebingen, ETH Zurich, and FU Berlin
// SPDX-License-Identifier: BSD-3-Clause
//
// --------------------------------------------------------------------------
// $Maintainer: Julianus Pfeuffer $
// $Authors: Julianus Pfeuffer $
// --------------------------------------------------------------------------
 
#pragma once
 
#include <OpenMS/DATASTRUCTURES/DefaultParamHandler.h>
#include <OpenMS/CONCEPT/Exception.h>
#include <OpenMS/CONCEPT/LogStream.h>
#include <OpenMS/KERNEL/ConsensusMap.h>
 
#include <algorithm>
#include <vector>
#include <set>
 
namespace OpenMS
{
 
  class OPENMS_DLLAPI IDScoreSwitcherAlgorithm:
    public DefaultParamHandler
  {
  public:
    IDScoreSwitcherAlgorithm ();
 
    enum class ScoreType
    {
      RAW,
      RAW_EVAL,
      PP,
      PEP,
      FDR,
      QVAL,
    };
 
    bool isScoreType(const String& score_name, const ScoreType& type)
    {
      String chopped = score_name;
      if (chopped.hasSuffix("_score"))
      {
        chopped = chopped.chop(6);
      }
      const std::set<String>& possible_types = type_to_str_[type];
      return possible_types.find(chopped) != possible_types.end();
    }
 
    static ScoreType getScoreType(String score_type)
    {
      if (score_type.hasSuffix("_score"))
      {
        score_type = score_type.chop(6);
      }
      score_type.toLower();     
      score_type.erase(std::remove_if(score_type.begin(), score_type.end(), 
                [](unsigned char c) { return c == '-' || c == '_' || c == ' '; }), 
                score_type.end());
 
      const std::map<String, ScoreType> s_to_type =
      {
        {"raw", ScoreType::RAW},
        {"rawevalue", ScoreType::RAW_EVAL},
        {"qvalue", ScoreType::QVAL},
        {"fdr", ScoreType::FDR},
        {"falsediscoveryrate", ScoreType::FDR},
        {"pep", ScoreType::PEP},
        {"posteriorerrorprobability", ScoreType::PEP},
        {"posteriorprobabilty", ScoreType::PP},
        {"pp", ScoreType::PP}
      };
 
      if (auto it = s_to_type.find(score_type); it != s_to_type.end())
      {
        return it->second;
      }
      else
      {
        throw Exception::MissingInformation(__FILE__, __LINE__,
                                            OPENMS_PRETTY_FUNCTION, String("Unknown score type ") + score_type);
      }
    }
 
    bool isScoreTypeHigherBetter(ScoreType score_type)
    {
      return type_to_better_[score_type];
    }
 
    /*
      * @brief Gets a vector of all score names that are used in OpenMS.
      *
      * @return A vector of all score names that are used in OpenMS (e.g., "q-value", "ln(hyperscore)").
    */
    std::vector<String> getScoreTypeNames();
 
    template <typename IDType>
    void switchScores(IDType& id, Size& counter)
    {
      for (auto hit_it = id.getHits().begin();
           hit_it != id.getHits().end(); ++hit_it, ++counter)
      {
        if (!hit_it->metaValueExists(new_score_))
        {
          std::stringstream msg;
          msg << "Meta value '" << new_score_ << "' not found for " << *hit_it;
          throw Exception::MissingInformation(__FILE__, __LINE__,
                                              OPENMS_PRETTY_FUNCTION, msg.str());
        }
 
        const String& old_score_meta = (old_score_.empty() ? id.getScoreType() :
                                 old_score_);
        const DataValue& dv = hit_it->getMetaValue(old_score_meta);
        if (!dv.isEmpty()) // meta value for old score already exists
        {
          // TODO: find a better way to check if old score type is something different (even if it has same name)
          // This currently, is a workaround for e.g., having Percolator_qvalue as meta value and same q-value as main score (getScore()).
          // Note by jpfeuffer: The problem with this is, that this may add the old score to some of the hits if different, but not
          // all, in case one is by chance the same. I would be fine with this, if it was done in the beginning and checked
          // for every score.
          if (fabs((double(dv) - hit_it->getScore()) * 2.0 /
                   (double(dv) + hit_it->getScore())) > tolerance_)
          {          
            hit_it->setMetaValue(old_score_meta + "~", hit_it->getScore());
          }
        }
        else
        {
          hit_it->setMetaValue(old_score_meta, hit_it->getScore());
        }
        hit_it->setScore(hit_it->getMetaValue(new_score_));
      }
      id.setScoreType(new_score_type_);
      id.setHigherScoreBetter(higher_better_);
    }
 
    template<class IDType>
    void switchToGeneralScoreType(std::vector<IDType>& id, ScoreType type, Size& counter)
    {
      if (id.empty()) return;
      String t = findScoreType(id[0], type);
      if (t.empty())
      {
        String msg = "First encountered ID does not have the requested score type.";
        throw Exception::MissingInformation(__FILE__, __LINE__,
                                            OPENMS_PRETTY_FUNCTION, msg);
      } 
      else if (t == id[0].getScoreType())
      {
        // we assume that all the other peptide ids
        // also already have the correct score set
        return;
      }
 
      if (t.hasSuffix("_score"))
      {
        new_score_type_ = t.chop(6);
      }
      else
      {
        new_score_type_ = t;
      }
      new_score_ = t;
 
      if (type != ScoreType::RAW && higher_better_ != type_to_better_[type])
      {
        OPENMS_LOG_WARN << "Requested non-raw score type does not match the expected score direction. Correcting!\n";
        higher_better_ = type_to_better_[type];
      }
      for (auto& i : id)
      {
        switchScores(i, counter);
      }
    }
 
    void switchToGeneralScoreType(ConsensusMap& cmap, ScoreType type, Size& counter, bool unassigned_peptides_too = true)
    {
      String new_type = "";
      for (const auto& f : cmap)
      {
        const auto& ids = f.getPeptideIdentifications();
        if (!ids.empty())
        {
          new_type = findScoreType(ids[0], type);
          if (new_type == ids[0].getScoreType())
          {
            return;
          }
          else
          {
            break;
          }
        }
      }
 
      if (new_type.empty())
      {
        String msg = "First encountered ID does not have the requested score type.";
        throw Exception::MissingInformation(__FILE__, __LINE__,
                                            OPENMS_PRETTY_FUNCTION, msg);
      }
 
      if (new_type.hasSuffix("_score"))
      {
        new_score_type_ = new_type.chop(6);
      }
      else
      {
        new_score_type_ = new_type;
      }
      new_score_ = new_type;
 
      if (type != ScoreType::RAW && higher_better_ != type_to_better_[type])
      {
        OPENMS_LOG_WARN << "Requested non-raw score type does not match the expected score direction. Correcting!\n";
        higher_better_ = type_to_better_[type];
      }
 
      const auto switchScoresSingle = [&counter,this](PeptideIdentification& id){switchScores(id,counter);};
      cmap.applyFunctionOnPeptideIDs(switchScoresSingle, unassigned_peptides_too);
    }
 
    
    template <typename IDType>
    String findScoreType(IDType& id, IDScoreSwitcherAlgorithm::ScoreType type)
    {
      const String& curr_score_type = id.getScoreType();
      const std::set<String>& possible_types = type_to_str_[type];
      if (possible_types.find(curr_score_type) != possible_types.end())
      {
        OPENMS_LOG_INFO << "Requested score type already set as main score: " + curr_score_type + "\n";
        return curr_score_type;
      }
      else
      {
        if (id.getHits().empty())
        {
          OPENMS_LOG_WARN << "Identification entry used to check for alternative score was empty.\n";
          return "";
        }
        const auto& hit = id.getHits()[0];
        for (const auto& poss_str : possible_types)
        {
          if (hit.metaValueExists(poss_str)) 
          {
            return poss_str;
          }
          else if (hit.metaValueExists(poss_str + "_score")) 
          {
            return poss_str + "_score";
          }
        }
        OPENMS_LOG_WARN << "Score of requested type not found in the UserParams of the checked ID object.\n";
        return "";
      }
    }
 
  private:
    void updateMembers_() override;
 
    const double tolerance_ = 1e-6;
 
    String new_score_, new_score_type_, old_score_;
 
    bool higher_better_; // for the new scores, are higher ones better?
 
    std::map<ScoreType, std::set<String>> type_to_str_ =
        {
            //TODO introduce real meaningful score names for XTandem, Mascot etc. (e.g., hyperscore)
            {ScoreType::RAW, {"svm", "MS:1001492", "XTandem", "OMSSA", "SEQUEST:xcorr", "Mascot", "mvh", "hyperscore", "ln(hyperscore)"}},
            //TODO find out reasonable raw scores for SES that provide E-Values as main score or see below
            //TODO there is no test for spectraST idXML, so I don't know its score
            //TODO check if we should combine RAW and RAW_EVAL:
            // What if a SE does not have an e-value score (spectrast, OMSSA, crux/sequest, myrimatch),
            // then you need additional if's/try's
            {ScoreType::RAW_EVAL, {"expect", "SpecEValue", "E-Value", "evalue", "MS:1002053", "MS:1002257"}},
            {ScoreType::PP, {"Posterior Probability"}},
            {ScoreType::PEP, {"Posterior Error Probability", "pep", "MS:1001493"}}, // TODO add CV terms
            {ScoreType::FDR, {"FDR", "fdr", "false discovery rate"}},
            {ScoreType::QVAL, {"q-value", "qvalue", "MS:1001491", "q-Value", "qval"}}
        };
 
    std::map<ScoreType, bool> type_to_better_ =
        {
            {ScoreType::RAW, true}, //TODO this might actually not always be true
            {ScoreType::RAW_EVAL, false},
            {ScoreType::PP, true},
            {ScoreType::PEP, false},
            {ScoreType::FDR, false},
            {ScoreType::QVAL, false}
        };        
  };
} // namespace OpenMS