libbpp-phyl-dev 2.0.3-1 / usr / include / Bpp / Phyl / Model / AbstractSubstitutionModel.h

//
// File: AbstractSubstitutionModel.h
// Created by: Julien Dutheil
// Created on: Tue May 27 10:31:49 2003
//

/*
   Copyright or © or Copr. Bio++ Development Team, (November 16, 2004)

   This software is a computer program whose purpose is to provide classes
   for phylogenetic data analysis.

   This software is governed by the CeCILL  license under French law and
   abiding by the rules of distribution of free software.  You can  use,
   modify and/ or redistribute the software under the terms of the CeCILL
   license as circulated by CEA, CNRS and INRIA at the following URL
   "http://www.cecill.info".

   As a counterpart to the access to the source code and  rights to copy,
   modify and redistribute granted by the license, users are provided only
   with a limited warranty  and the software's author,  the holder of the
   economic rights,  and the successive licensors  have only  limited
   liability.

   In this respect, the user's attention is drawn to the risks associated
   with loading,  using,  modifying and/or developing or reproducing the
   software by the user in light of its specific status of free software,
   that may mean  that it is complicated to manipulate,  and  that  also
   therefore means  that it is reserved for developers  and  experienced
   professionals having in-depth computer knowledge. Users are therefore
   encouraged to load and test the software's suitability as regards their
   requirements in conditions enabling the security of their systems and/or
   data to be ensured and,  more generally, to use and operate it in the
   same conditions as regards security.

   The fact that you are presently reading this means that you have had
   knowledge of the CeCILL license and that you accept its terms.
 */

#ifndef _ABSTRACTSUBSTITUTIONMODEL_H_
#define _ABSTRACTSUBSTITUTIONMODEL_H_

#include "SubstitutionModel.h"

#include <Bpp/Numeric/AbstractParameterAliasable.h>
#include <Bpp/Numeric/VectorTools.h>

namespace bpp
{
/**
 * @brief Partial implementation of the SubstitutionModel interface.
 *
 * This abstract class provides some fields, namely:
 * - alphabet_: a pointer toward the alphabet,
 * - size_: the size of the alphabet, a parameter frequently called during various computations,
 * - rate_: the rate of the model
 * - generator_, leftEigenVectors_, rightEigenVectors_: useful matrices,
 * - eigenValues_, iEigenValues_, freq_: useful vectors.
 * - isDiagonalizable_ : boolean value useful for computation of the exponential
 *
 * Access methods for these fields are implemented.
 *
 * This class also provides the updateMatrices() method, which computes eigen values and vectors and fills the corresponding vector (eigenValues_)
 * and matrices (leftEigenVectors_ and rightEigenVectors_) from the generator.
 *
 * The freq_ vector and generator_ matrices are hence the only things to provide to
 * create a substitution model.
 * It is also possible to redefine one of these methods for better efficiency.
 * The Pij_t, dPij_dt and d2Pij_dt2 are particularly inefficient since the matrix formula
 * is used to compute all probabilities, and then the result for the initial and final state
 * of interest is retrieved.
 *
 * @note This class is dedicated to "simple" substitution models, for which the number of states is equivalent to the number of characters in the alphabet.
 * Consider using the MarkovModulatedSubstitutionModel for more complexe cases.
 */
class AbstractSubstitutionModel :
  public virtual SubstitutionModel,
  public virtual AbstractParameterAliasable
{
protected:
  /**
   * @brief The alphabet relevant to this model.
   */
  const Alphabet* alphabet_;

  /**
   * @brief The size of the generator, i.e. the number of states.
   */
  unsigned int size_;

  /**
   * @brief The rate of the model (default: 1). The generator (and all
   * its vectorial components) is independent of the rate, since it
   * should be normalized.
   */
  
  double rate_;

  /**
   * @brief The list of supported chars.
   */
  std::vector<int> chars_;

  /**
   * @brief The generator matrix \f$Q\f$ of the model.
   */
  RowMatrix<double> generator_;

  /**
   * @brief These ones are for bookkeeping:
   */
  mutable RowMatrix<double> pijt_;
  mutable RowMatrix<double> dpijt_;
  mutable RowMatrix<double> d2pijt_;

  /**
   * @brief The \f$U\f$ matrix made of left eigen vectors (by row).
   */
  RowMatrix<double> leftEigenVectors_;

  /**
   * @brief The \f$U^-1\f$ matrix made of right eigen vectors (by column).
   */
  RowMatrix<double> rightEigenVectors_;

  /**
   * @brief The vector of eigen values.
   */
  Vdouble eigenValues_;

  /**
   * @brief The vector of the imaginary part of the eigen values.
   */
  Vdouble iEigenValues_;

  /**
   * @brief boolean value for diagonalizability in R of the generator_
   */
  
  bool isDiagonalizable_;

  /**
   * @brief The vector of equilibrium frequencies.
   */
  Vdouble freq_;

  /**
   * @brief Tell if the eigen decomposition should be performed.
   */
  bool eigenDecompose_;

public:
  AbstractSubstitutionModel(const Alphabet* alpha, const std::string& prefix);

  AbstractSubstitutionModel(const AbstractSubstitutionModel& model) :
    AbstractParameterAliasable(model),
    alphabet_(model.alphabet_),
    size_(model.size_),
    rate_(model.rate_),
    chars_(model.chars_),
    generator_(model.generator_),
    pijt_(model.pijt_),
    dpijt_(model.dpijt_),
    d2pijt_(model.d2pijt_),
    leftEigenVectors_(model.leftEigenVectors_),
    rightEigenVectors_(model.rightEigenVectors_),
    eigenValues_(model.eigenValues_),
    iEigenValues_(model.iEigenValues_),
    isDiagonalizable_(model.isDiagonalizable_),
    freq_(model.freq_),
    eigenDecompose_(model.eigenDecompose_)
  {}

  AbstractSubstitutionModel& operator=(const AbstractSubstitutionModel& model)
  {
    AbstractParameterAliasable::operator=(model);
    alphabet_          = model.alphabet_;
    size_              = model.size_;
    rate_              = model.rate_;
    chars_             = model.chars_;
    generator_         = model.generator_;
    pijt_              = model.pijt_;
    dpijt_             = model.dpijt_;
    d2pijt_            = model.d2pijt_;
    leftEigenVectors_  = model.leftEigenVectors_;
    rightEigenVectors_ = model.rightEigenVectors_;
    eigenValues_       = model.eigenValues_;
    iEigenValues_      = model.iEigenValues_;
    isDiagonalizable_  = model.isDiagonalizable_;
    freq_              = model.freq_;
    eigenDecompose_    = model.eigenDecompose_;
    return *this;
  }
  virtual ~AbstractSubstitutionModel() {}

#ifndef NO_VIRTUAL_COV
  virtual AbstractSubstitutionModel* clone() const = 0;
#endif

public:
  const Alphabet* getAlphabet() const { return alphabet_; }

  const std::vector<int>& getAlphabetChars() const { return chars_; }

  int getAlphabetChar(unsigned int i) const { return chars_[i]; }

  std::vector<unsigned int> getModelStates(int i) const { return VectorTools::whichAll(chars_, i); }

  virtual const Vdouble& getFrequencies() const { return freq_; }

  const Matrix<double>& getGenerator() const { return generator_; }

  virtual const Matrix<double>& getPij_t(double t) const;
  virtual const Matrix<double>& getdPij_dt(double t) const;
  virtual const Matrix<double>& getd2Pij_dt2(double t) const;

  const Vdouble& getEigenValues() const { return eigenValues_; }

  const Vdouble& getIEigenValues() const { return iEigenValues_; }

  bool isDiagonalizable() const { return isDiagonalizable_;}
  
  const Matrix<double>& getRowLeftEigenVectors() const { return leftEigenVectors_; }
  const Matrix<double>& getColumnRightEigenVectors() const { return rightEigenVectors_; }

  virtual double freq(unsigned int i) const { return freq_[i]; }

  virtual double Qij(unsigned int i, unsigned int j) const { return generator_(i, j); }

  virtual double Pij_t    (unsigned int i, unsigned int j, double t) const { return getPij_t(t) (i, j); }
  virtual double dPij_dt  (unsigned int i, unsigned int j, double t) const { return getdPij_dt(t) (i, j); }
  virtual double d2Pij_dt2(unsigned int i, unsigned int j, double t) const { return getd2Pij_dt2(t) (i, j); }

  double getInitValue(unsigned int i, int state) const throw (BadIntException);

  void setFreqFromData(const SequenceContainer& data, unsigned int pseudoCount = 0);

  virtual void setFreq(std::map<int, double>&);

  void enableEigenDecomposition(bool yn) { eigenDecompose_ = yn; }

  bool enableEigenDecomposition() { return eigenDecompose_; }

  /**
   * @brief Tells the model that a parameter value has changed.
   *
   * This updates the matrices consequently.
   */
  virtual void fireParameterChanged(const ParameterList& parameters)
  {
    AbstractParameterAliasable::fireParameterChanged(parameters);
    if ((parameters.size()!=1) || (parameters[0].getName()!=getNamespace()+"rate"))
      updateMatrices();
  }

  void addRateParameter();

protected:
  /**
   * @brief Diagonalize the \f$Q\f$ matrix, and fill the eigenValues_, iEigenValues_, 
   * leftEigenVectors_ and rightEigenVectors_ matrices.
   *
   * The generator_ matrix and freq_ vector must be initialized.
   *
   * Eigen values and vectors are computed from the generator and
   * assigned to the eigenValues_ for the real part, iEigenValues_ for
   * the imaginary part, rightEigenVectors_ and leftEigenVectors_
   * variables. isDiagonalizable_ checks if the generator_ is
   * diagonalizable in R.
   *
   * The optional rate parameter is not taken into account in this
   * method to prevent unnecessary computation.
   */
  virtual void updateMatrices();

public:
  /**
   * @brief Get the scalar product of diagonal elements of the generator
   * and the frequencies vector.
   * If the generator is normalized, then scale=1. Otherwise each element
   * must be multiplied by 1/scale.
   *
   * @return Minus the scalar product of diagonal elements and the frequencies vector.
   */
  double getScale() const;

  /**
   * @brief Set the rate of the generator, defined as the scalar
   * product of diagonal elements of the generator and the frequencies
   * vector.
   *
   * When the generator is normalized, scale=1. Otherwise each element
   * is multiplied such that the correct scale is set.
   *
   */
  void setScale(double scale);

  /**
   * @brief Get the rate of the model
   */
  
  virtual double getRate() const;

  /**
   * @brief Set the rate of the model (must be positive).
   * @param rate must be positive.
   */
  
  virtual void setRate(double rate);


  friend class AbstractBiblioSubstitutionModel;

};


/**
 * @brief Partial implementation of the ReversibleSubstitutionModel interface.
 *
 * This abstract class adds the exchangeability_ fields to the AbstractSubstitutionModel class.
 * Access methods for this field is implemented.
 *
 * This class also overrides the updateMatrices() method, which
 * updates the generator_ matrix from the exchangeability_ matrix and
 * freq_ vector. It then computes eigen values and vectors and fills
 * the corresponding vector (eigenValues_) and matrices
 * (leftEigenVectors_ and rightEigenVectors_). Because of
 * reversibility, isDiagonalizable_ is set to true.
 *
 * The freq_ vector and exchangeability_ matrices are hence the only things to provide to
 * create a substitution model.
 * It is also possible to redefine one of these methods for better efficiency.
 * The Pij_t, dPij_dt and d2Pij_dt2 are particularly inefficient since the matrix formula
 * is used to compute all probabilities, and then the result for the initial and final state
 * of interest is retrieved.
 *
 * @note This class is dedicated to "simple" substitution models, for which the number of states is equivalent to the number of characters in the alphabet.
 * Consider using the MarkovModulatedSubstitutionModel for more complexe cases.
 */
class AbstractReversibleSubstitutionModel :
  public virtual AbstractSubstitutionModel,
  public virtual ReversibleSubstitutionModel
{
protected:
  /**
   * @brief The exchangeability matrix \f$S\f$ of the model.
   */
  RowMatrix<double> exchangeability_;

public:
  AbstractReversibleSubstitutionModel(const Alphabet* alpha, const std::string& prefix) :
    AbstractParameterAliasable(prefix),
    AbstractSubstitutionModel(alpha, prefix),
    exchangeability_(size_, size_) {
    isDiagonalizable_=true;
  }

  virtual ~AbstractReversibleSubstitutionModel() {}

#ifndef NO_VIRTUAL_COV
  virtual AbstractReversibleSubstitutionModel* clone() const = 0;
#endif


public:
  const Matrix<double>& getExchangeabilityMatrix() const { return exchangeability_; }
  double Sij(unsigned int i, unsigned int j) const { return exchangeability_(i, j); }

protected:
  /**
   * @brief Compute and diagonalize the \f$Q\f$ matrix, and fill the eigenValues_,
   * leftEigenVectors_ and rightEigenVectors_ matrices.
   *
   * The exchangeability_ matrix and freq_ vector must be initialized.
   * This function computes the generator_ matrix with the formula
   * \f[
   * Q = S \times \pi
   * \f]
   * where \f$Q\f$ is the generator matrix, \f$S\f$ is the exchangeability matrix and
   * \f$Pi\f$ the diagonal matrix with frequencies.
   *
   * The generator is then scaled so that
   * \f[
   * \sum_i Q_{i,i} \times \pi_i = -1
   * \f]
   * (\f$\pi_i\f$ are the equilibrium frequencies).
   *
   * Eigen values and vectors are computed from the scaled generator and assigned to the
   * eigenValues_, rightEigenVectors_ and leftEigenVectors_ variables.
   */
  virtual void updateMatrices();
};

} //end of namespace bpp.

#endif  //_ABSTRACTSUBSTITUTIONMODEL_H_