R/RcppExports.R

In TomMayo/kmermods: Represents Kmers As Integers Along The Genome

# This file was generated by Rcpp::compileAttributes
# Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393

#' Finds the integer that represents the nucleotide
#'
#' Returns an integer for {G,A,T,G} and NA for 'N'
#' 
#' @param letter A single nucleotide, as a character.
#' @param alph_vect The alphabet we are using. A dataframe created using 
#' build_alphabet()
#' @return An integer, or NA if the input is 'N'
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
let2base_c <- function(letter, alph_vect) {
    .Call('kmermods_let2base_c', PACKAGE = 'kmermods', letter, alph_vect)
}

#' Converts a base representation (base 4 default) to normal base 10
#'
#' Returns an integer in base 10, calculated from the base 4 (default) 
#' representation
#' 
#' @inheritParams base2kmer
#' @inheritParams kmer2base
#' @return An integer, corresponding to the number in base 4(default)
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
base5to10_c <- function(number, k, base = 5L) {
    .Call('kmermods_base5to10_c', PACKAGE = 'kmermods', number, k, base)
}

#' Converts a normal (base 10) integer to base representation (base 4 default) 
#'
#' Returns a base 4 (default) number as a vector calculated from the  base 10 
#' equivalent
#' 
#' @param number_b10 An integer in base 10, representing a kmer
#' @inheritParams base2kmer
#' @inheritParams kmer2base
#' @return A vector of integers representing a number in base 4 (default)
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
convert10to5_c <- function(number_b10, k, base = 4L) {
    .Call('kmermods_convert10to5_c', PACKAGE = 'kmermods', number_b10, k, base)
}

#' Converts a normal (base 10) integer to base representation (base 4 default) 
#'
#' Returns a base 4 (default) number as a vector calculated from the  base 10 
#' equivalent. Same as convert10to5_c but slightly faster
#' 
#' @param number_b10 An integer in base 10, representing a kmer
#' @inheritParams base2kmer
#' @inheritParams kmer2base
#' @return A vector of integers representing a number in base 4 (default)
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
convert10tobase_c <- function(number_b10, k, base = 4L) {
    .Call('kmermods_convert10tobase_c', PACKAGE = 'kmermods', number_b10, k, base)
}

#' Counts the number of mismatches between two kmers, when represented in 
#' integer format 
#'
#' Returns a count of the number of mismatches between two kmers (of equal 
#' length), which are represented in integer format. For instance the two kmers
#' AAA and ATA have one mismatch. This function bypasses the conversion to 
#' strings.
#' 
#' @param kmer_1 An integer in base 10, representing a kmer
#' @param kmer_2 An integer in base 10, representing a kmer
#' @inheritParams base2kmer
#' @inheritParams kmer2base
#' @return A count of the mismatches
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
mismatch_kmers <- function(kmer_1, kmer_2, k, base = 4L) {
    .Call('kmermods_mismatch_kmers', PACKAGE = 'kmermods', kmer_1, kmer_2, k, base)
}

#' Creates the update for the parameter vector according to the error term 
#' and the kmers involved
#'
#' Returns a vector of length equal to the parameter vector, which is update 
#' vector to be added to the parameter vector
#' 
#' @param kmers A vector of integers representing the kmers that were 
#' present leading to the error term
#' @param update_vec The parameter update vector so far
#' @param err_term The error term used for the update
#' @param err_term The error term used for the update
#' @return add_to_update TRUE/FALSE, whether the calculated update vector for 
#' the run is added automatically to the input update vector, so that the 
#' function returns a running total, which is the faster option.
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
update_paras <- function(kmers, update_vec, err_term, add_to_update = TRUE) {
    .Call('kmermods_update_paras', PACKAGE = 'kmermods', kmers, update_vec, err_term, add_to_update)
}

#' Computes the dot product of the parameters with the kmer counts, with or 
#' without it being warped
#'
#' Computes the dot product of the parameters with the kmer counts, with or 
#' without it being warped
#' 
#' @param kmers is a vector of integers of any length representing kmers over
#'  a window
#' @param paras is a vector of length equal to the total number of kmers
#' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry 
#' @return A double, representing the dot product of the parameters with what 
#' would usually be the vector of kmer abundances (warped or not)
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @examples 
#' kmers <- c(0,7,89,45,75,65,22,12)
#' paras <- rep(1,100)
#' kmer_dot_prod(kmers,paras)
#' @export
kmer_dot_prod_c <- function(indices, params, warp_ = NULL) {
    .Call('kmermods_kmer_dot_prod_c', PACKAGE = 'kmermods', indices, params, warp_)
}

#' Takes a longer kmer and converts it into a vector of its shorter kmers
#'
#' Takes in a single long kmer, for example a 26mer, and returns the composite 
#' shorter kmers for some k, for example, all the 10mers within the 26mer. 
#' It is designed to be used to unwrap a dense representation of the kmers: 
#' for example, given the 26mers every 19 base pairs, we can quickly extract 
#' all 8mers.
#' 
#' @param kmer A number representing a single large kmer, with k = old_len
#' @param old_len The length of the kmer represented by kmer (ie k)
#' @param new_len The length of the kmers we want to extract, eg, if we want 
#' 8mers, new_len = 8
#' @param base The length of the alphabet. For normal DNA sequence this is 4.
#' @param num_kmers The number of the new kmers to return. The functions 
#' returns the first num_kmers in the vector. Default value is 0, which returns 
#' all kmers.
#' @return A vector of doubles (actually integers represented as doubles), 
#' representing all the (new_len)-mers contained within the single (old_len)-mer.
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
unwrap_kmers <- function(kmer, old_len = 26, new_len = 8, base = 4, num_kmers = 0L) {
    .Call('kmermods_unwrap_kmers', PACKAGE = 'kmermods', kmer, old_len, new_len, base, num_kmers)
}

#' Takes in a single long kmer, for example a 26mer, and returns the composite 
#' shorter kmers for some k, for example, all the 10mers within the 26mer. 
#' It is designed to be used to unwrap a dense representation of the kmers: 
#' for example, given the 26mers every 19 base pairs, we can quickly extract 
#' all 8mers.
#' 
#' @param kmer A number representing a single large kmer, with k = old_len
#' @param old_len The length of the kmer represented by kmer (ie k)
#' @param new_len The length of the kmers we want to extract, eg, if we want 
#' 8mers, new_len = 8
#' @param base The length of the alphabet. For normal DNA sequence this is 4.
#' @param num_kmers The number of the new kmers to return. The functions 
#' returns the first num_kmers in the vector. Default value is 0, which returns 
#' all kmers.
#' @return A vector of doubles (actually integers represented as doubles), 
#' representing all the (new_len)-mers contained within the single (old_len)-mer.
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
unwrap_kmers_vect <- function(kmers, old_len = 26, new_len = 8, base = 4, num_kmers = 0L) {
    .Call('kmermods_unwrap_kmers_vect', PACKAGE = 'kmermods', kmers, old_len, new_len, base, num_kmers)
}

#' Computes the vector with which to update the parameters in a logistic regression
#' onto the peaks
#'
#' This calculates the prediction for each region, in a non-sliding scheme, 
#' calculates the error and returns the update vector for all of the regions.
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param peaks is a matrix giving the locations of the peaks on the chromosome,
#' the first column is starts, second is ends, inclusive, indexed from 1
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' //' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return A vector, representing the amount to update the parameter vector
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
params_peaks_noslide <- function(kmers_win, params, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_params_peaks_noslide', PACKAGE = 'kmermods', kmers_win, params, peaks, win_size, chrom_loc, warp_)
}

#' Computes the total error of the predictions.
#'
#' This calculates the total of the absolute values of the errors, given a set
#' of parameters and outputs, over a given region.
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param peaks is a matrix giving the locations of the peaks on the chromosome,
#' the first column is starts, second is ends, inclusive, indexed from 1
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return The total of the absolute errors, with a breakdown
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
total_error <- function(kmers_win, params, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_total_error', PACKAGE = 'kmermods', kmers_win, params, peaks, win_size, chrom_loc, warp_)
}

#' Computes the vector with which to update the parameters in a logistic regression
#' onto the peaks and also returns error calculations
#'
#' This calculates the prediction for each region, in a non-sliding scheme, 
#' calculates the error and returns the update vector for all of the regions.
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param peaks is a matrix giving the locations of the peaks on the chromosome,
#' the first column is starts, second is ends, inclusive, indexed from 1
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' //' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return A list, the first element is a vector representing the amount to 
#' update the parameter vector, the second is the sum of errors of different 
#' types
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
params_peaks_noslide_w_error <- function(kmers_win, params, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_params_peaks_noslide_w_error', PACKAGE = 'kmermods', kmers_win, params, peaks, win_size, chrom_loc, warp_)
}

#' Computes the predictions for the logistic regression
#'
#' This calculates the predictions at each window, returning a probability of 
#' seeing a peak there
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return A vector of probabilities
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
predict_peaks <- function(kmers_win, params, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_predict_peaks', PACKAGE = 'kmermods', kmers_win, params, peaks, win_size, chrom_loc, warp_)
}

#' L1 regulatisation proximal operator
#'
#' This function computes the proximal operator for L1- regularised regression
#' (lasso) and returns the new vector.
#' 
#' @param params is a vector of length equal to the total number of kmers, 
#' representing the parameters in the model
#' @param thresh is the threshold for the proximal operator for l1 regularised
#' regression
#' @return A vector of parameters
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
l1_prox_op <- function(params, thresh) {
    .Call('kmermods_l1_prox_op', PACKAGE = 'kmermods', params, thresh)
}

#' Calculates the log likelihood of the data, given some parameters for a 
#' logistic regression
#'
#' Given the kmers, set of peaks and parameters, this returns the log likelihood.
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param peaks is a matrix giving the locations of the peaks on the chromosome,
#' the first column is starts, second is ends, inclusive, indexed from 1
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' //' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return A double, representing the log likelihood
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
loglik_logreg <- function(kmers_win, params, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_loglik_logreg', PACKAGE = 'kmermods', kmers_win, params, peaks, win_size, chrom_loc, warp_)
}

#' Computes the gradient of the log likelihood of a logistic regression
#'
#' This calculates the prediction for each region, in a non-sliding scheme, 
#' calculates the error and returns the update vector for all of the regions.
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param peaks is a matrix giving the locations of the peaks on the chromosome,
#' the first column is starts, second is ends, inclusive, indexed from 1
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' //' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return A vector, representing the gradient (direction to update the 
#' parameter vector)
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
loglik_log_reg_grad <- function(kmers_win, params, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_loglik_log_reg_grad', PACKAGE = 'kmermods', kmers_win, params, peaks, win_size, chrom_loc, warp_)
}

#' Computes the gradient of the log likelihood of a logistic regression
#'
#' This calculates the prediction for each region, in a non-sliding scheme, 
#' calculates the error and returns the update vector for all of the regions.
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param peaks is a matrix giving the locations of the peaks on the chromosome,
#' the first column is starts, second is ends, inclusive, indexed from 1
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' //' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return A vector, representing the gradient (direction to update the 
#' parameter vector)
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
log_reg_wrapper <- function(kmers_win, params, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_log_reg_wrapper', PACKAGE = 'kmermods', kmers_win, params, peaks, win_size, chrom_loc, warp_)
}

#' Calculates the log likelihood of the data, given some parameters for a 
#' logistic regression
#'
#' Given the kmers, set of peaks and parameters, this returns the log likelihood.
#' 
#' @param kmers_win is a vector of integers of any length representing kmers in
#' a region
#' @param paras is a vector of length equal to the total number of kmers
#' @param grad is a vector of length equal to the total number of kmers, 
#' representing the gradient of the log likelihood at that point
#' @param alphas is a vector of alphas, giving us the values for the line 
#' search, this will test loglik at params + alpha*grad for each alpha
#' @param peaks is a matrix giving the locations of the peaks on the chromosome,
#' the first column is starts, second is ends, inclusive, indexed from 1
#' @param win_size is the length of the sliding window we are using
#' @param chrom_loc is the position of the first kmer along the chromosome - 
#' this avoids indexing errors when splitting up the data
#' //' @param warp is a vector of length as long as the kmer vector, with the 
#' multiplicative weights for how much to warp the entry
#' @return A double, representing the log likelihood
#' @author Tom Mayo \email{t.mayo@@ed.ac.uk}
#' @export
loglik_linesearch <- function(kmers_win, params, grad, alphas, peaks, win_size, chrom_loc, warp_ = NULL) {
    .Call('kmermods_loglik_linesearch', PACKAGE = 'kmermods', kmers_win, params, grad, alphas, peaks, win_size, chrom_loc, warp_)
}