SimplePca.hpp

Go to the documentation of this file.

#ifndef SCRAN_SIMPLE_PCA_HPP
#define SCRAN_SIMPLE_PCA_HPP
 
#include "../utils/macros.hpp"
 
#include "tatami/tatami.hpp"
#include "irlba/irlba.hpp"
#include "Eigen/Dense"
 
#include <vector>
#include <cmath>
 
#include "convert.hpp"
#include "utils.hpp"
 
namespace scran {
 
class SimplePca {
public:
    struct Defaults {
        static constexpr int rank = 10;
 
        static constexpr bool scale = false;
 
        static constexpr bool transpose = true;
 
        static constexpr int num_threads = 1;
 
        static constexpr bool return_rotation = false;
 
        static constexpr bool return_center = false;
 
        static constexpr bool return_scale = false;
    };
private:
    bool scale = Defaults::scale;
    bool transpose = Defaults::transpose;
    int rank = Defaults::rank;
 
    bool return_rotation = Defaults::return_rotation;
    bool return_center = Defaults::return_center;
    bool return_scale = Defaults::return_scale;
 
    int nthreads = Defaults::num_threads;
 
public:
    SimplePca& set_rank(int r = Defaults::rank) {
        rank = r;
        return *this;
    }
 
    SimplePca& set_scale(bool s = Defaults::scale) {
        scale = s;
        return *this;
    }
 
    SimplePca& set_transpose(bool t = Defaults::transpose) {
        transpose = t;
        return *this;
    }
 
    SimplePca& set_return_rotation(bool r = Defaults::return_rotation) {
        return_rotation = r;
        return *this;
    }
 
    SimplePca& set_return_center(bool r = Defaults::return_center) {
        return_center = r;
        return *this;
    }
 
    SimplePca& set_return_scale(bool r = Defaults::return_scale) {
        return_scale = r;
        return *this;
    }
 
    SimplePca& set_num_threads(int n = Defaults::num_threads) {
        nthreads = n;
        return *this;
    }
 
private:
    template<typename Data_, typename Index_>
    void run_sparse(
        const tatami::Matrix<Data_, Index_>* mat, 
        const irlba::Irlba& irb, 
        Eigen::MatrixXd& pcs, 
        Eigen::MatrixXd& rotation, 
        Eigen::VectorXd& variance_explained,
        Eigen::VectorXd& center_v,
        Eigen::VectorXd& scale_v,
        double& total_var)
    const {
        auto extracted = pca_utils::extract_sparse_for_pca(mat, nthreads); // row-major extraction to create a CSR matrix.
        pca_utils::SparseMatrix emat(mat->ncol(), mat->nrow(), std::move(extracted.values), std::move(extracted.indices), std::move(extracted.ptrs), nthreads); // CSC with genes in columns.
 
        size_t ngenes = emat.cols();
        center_v.resize(ngenes);
        scale_v.resize(ngenes);
        pca_utils::compute_mean_and_variance_from_sparse_matrix(emat, center_v, scale_v, nthreads);
        total_var = pca_utils::process_scale_vector(scale, scale_v);
 
        irlba::Centered<decltype(emat)> centered(&emat, &center_v);
        if (scale) {
            irlba::Scaled<decltype(centered)> scaled(&centered, &scale_v);
            irb.run(scaled, pcs, rotation, variance_explained);
        } else {
            irb.run(centered, pcs, rotation, variance_explained);
        }
    }
 
    template<typename Data_, typename Index_>
    void run_dense(
        const tatami::Matrix<Data_, Index_>* mat, 
        const irlba::Irlba& irb, 
        Eigen::MatrixXd& pcs, 
        Eigen::MatrixXd& rotation, 
        Eigen::VectorXd& variance_explained, 
        Eigen::VectorXd& center_v,
        Eigen::VectorXd& scale_v,
        double& total_var) 
    const {
        auto emat = pca_utils::extract_dense_for_pca(mat, nthreads); // get a column-major matrix with genes in columns.
 
        size_t ngenes = emat.cols();
        center_v.resize(ngenes);
        scale_v.resize(ngenes);
        pca_utils::compute_mean_and_variance_from_dense_matrix(emat, center_v, scale_v, nthreads);
        total_var = pca_utils::process_scale_vector(scale, scale_v);
 
        // Applying the centering and scaling now so we can do the PCA without any wrappers.
        pca_utils::apply_center_and_scale_to_dense_matrix(emat, center_v, scale, scale_v, nthreads);
        irb.run(emat, pcs, rotation, variance_explained); 
    }
 
    template<typename Data_, typename Index_>
    void run_internal(
        const tatami::Matrix<Data_, Index_>* mat, 
        Eigen::MatrixXd& pcs, 
        Eigen::MatrixXd& rotation, 
        Eigen::VectorXd& variance_explained, 
        Eigen::VectorXd& center_v,
        Eigen::VectorXd& scale_v,
        double& total_var) 
    const {
        irlba::EigenThreadScope t(nthreads);
        irlba::Irlba irb;
        irb.set_number(rank);
        irb.set_cap_number(true);
 
        if (mat->sparse()) {
            run_sparse(mat, irb, pcs, rotation, variance_explained, center_v, scale_v, total_var);
        } else {
            run_dense(mat, irb, pcs, rotation, variance_explained, center_v, scale_v, total_var);
        }
 
        pca_utils::clean_up(mat->ncol(), pcs, variance_explained);
        if (transpose) {
            pcs.adjointInPlace();
        }
    }
 
public:
    struct Results {
        Eigen::MatrixXd pcs;
 
        Eigen::VectorXd variance_explained;
 
        double total_variance = 0;
 
        Eigen::MatrixXd rotation;
 
        Eigen::VectorXd center;
 
        Eigen::VectorXd scale;
    };
 
    template<typename T, typename IDX>
    Results run(const tatami::Matrix<T, IDX>* mat) const {
        Results output;
 
        Eigen::MatrixXd rotation;
        Eigen::VectorXd center, scale;
        run_internal(mat, output.pcs, rotation, output.variance_explained, center, scale, output.total_variance);
 
        // Shifting them if we want to keep them.
        if (return_rotation) {
            output.rotation = std::move(rotation);
        }
        if (return_center) {
            output.center = std::move(center);
        }
        if (return_scale) {
            output.scale = std::move(scale);
        }
 
        return output;
    }
 
    template<typename T, typename IDX, typename X>
    Results run(const tatami::Matrix<T, IDX>* mat, const X* features) const {
        if (!features) {
            return run(mat);
        } else {
            auto subsetted = pca_utils::subset_matrix_by_features(mat, features);
            return run(subsetted.get());
        }
    }
};
 
}
 
#endif