Homec4science
Diffusion scATAC-seq (master) Browse

scATAC-seq/src/Clustering/ConsensusSequenceModelComputer.hpp
269172e8f4efmaster

Last Change
Raw File

ConsensusSequenceModelComputer.hpp
View Options

#ifndef CONSENSUSSEQUENCEMODELCOMPUTER_HPP
#define CONSENSUSSEQUENCEMODELCOMPUTER_HPP
#include <ModelComputer.hpp>
#include <Matrix2D.hpp>
#include <Matrix4D.hpp>
#include <ThreadPool.hpp>
/*!
* \brief The ConsensusSequenceModelComputer class provides the
* bases to compute the DNA sequence models, as computed
* by one of the expectation maximization algorithm for
* consensus DNA sequence data, given a data matrix and a partitition.
* The data partition must be a probability matrix, as returned
* by one of the classes deriving EMBase.
* The data must be a 3D matrix of dimensions NxLx4
* containing the probability of each base at each of the L different
* positions in N different regions.
* The models returned are the weighted aggregations
* of the consensus sequences, assigned to each class.
* The format of the model matrix returned is :
* 1) 4 * the number of classes
* 2) the model length + 1.
* The rows should by read 4 by 4 to get each class
* model. For instance, if the partition contains 3
* classes, the model matrix has 12 rows. The 1st
* class is contained in rows 1-4, the 2nd class in
* rows 5-8 and the 3rd class in rows 9-12.
* The 1st column always contains the class probabilities. The signal
* starts at the 2nd column. For each class, the 1st column sums
* up to 1.0
*/
class ConsensusSequenceModelComputer : public ModelComputer
{
public:
/*!
* \brief Constructs an object to retrieve
* the (consensus) sequence model given the data
* and their classification results.
* \param data the data.
* Its dimensions are :
* 1) the number of consensus sequences
* 2) the length of each consensus sequence
* 3) 4 for A, C, G, T
* \param post_prob the data class assignment
* probabilities. Its dimensions are :
* 1) the 1st dimension of the data matrix
* 2) the number of classes
* 3) the number of shift states
* 4) the number of flip states
* \param bckg_class whether the last class of the
* classification (posterior probabilities) is a
* background class.
* \param n_threads the number of parallel threads
* to run the computations. 0 means no parallel
* computing, everything is run on the main thread.
*/
ConsensusSequenceModelComputer(Matrix3D<double>&& data,
const Matrix4D<double>& post_prob,
bool bckg_class,
size_t n_threads) ;
/*!
* \brief Destructor.
*/
virtual ~ConsensusSequenceModelComputer() override ;
protected:
/*!
* \brief the threads.
*/
ThreadPool* threads ;
} ;
#endif // CONSENSUSSEQUENCEMODELCOMPUTER_HPP
c4science · Help