CSIDE.R

#' Runs CSIDE on a \code{\linkS4class{RCTD}} object with a single explanatory variable
#'
#' Identifies cell type specific differential expression (DE) as a function of the explanatory variable.
#' The design matrix contains an intercept column and a column of the explanatory variable. Uses maximum
#' likelihood estimation to estimate DE and standard errors for each gene and each cell type. Selects
#' genes with significant nonzero DE.
#'
#' @param myRCTD an \code{\linkS4class{RCTD}} object with annotated cell types e.g. from the \code{\link{run.RCTD}} function.
#' @param explanatory.variable a named numeric vector representing the explanatory variable used for explaining differential expression in CSIDE. Names of the variable
#' are the \code{\linkS4class{SpatialRNA}} pixel names, and values should be standardized between 0 and 1.
#' @param cell_types the cell types used for CSIDE. If null, cell types will be chosen with aggregate occurrences of
#' at least `cell_type_threshold`, as aggregated by \code{\link{aggregate_cell_types}}
#' @param cell_type_threshold (default 125) min occurrence of number of cells for each cell type to be used, as aggregated by \code{\link{aggregate_cell_types}}
#' @param gene_threshold (default 5e-5) minimum average normalized expression required for selecting genes
#' @param doublet_mode (default TRUE) if TRUE, uses RCTD doublet mode weights. Otherwise, uses RCTD full mode weights
#' @param sigma_gene (default TRUE) if TRUE, fits gene specific overdispersion parameter. If FALSE, overdispersion parameter is same across all genes.
#' @param weight_threshold (default NULL) the threshold of total normalized weights across all cell types
#' in \code{cell_types} per pixel to be included in the model. Default 0.99 for doublet_mode or 0.8 for full_mode.
#' @param PRECISION.THRESHOLD (default 0.05) for checking for convergence, the maximum parameter change per algorithm step
#' @param cell_types_present cell types (a superset of `cell_types`) to be considered as occurring often enough
#' to consider for gene expression contamination during the step filtering out marker genes of other cell types.
#' @param fdr (default 0.01) false discovery rate for hypothesis testing
#' @param test_genes_sig (default TRUE) logical controlling whether genes will be tested for significance
#' @param normalize_expr (default FALSE) if TRUE, constrains total gene expression to sum to 1 in each condition.
#' @param logs (default FALSE) if TRUE, writes progress to logs/de_logs.txt
#' @param test_error (default FALSE) if TRUE, exits after testing for error messages without running CSIDE.
#' If set to TRUE, this can be used to quickly evaluate if CSIDE will run without error.
#' @param log_fc_thresh (default 0.4) the natural log fold change cutoff for differential expression
#' @param fdr_method (default BH) if BH, uses the Benjamini-Hochberg method. Otherwise, uses local fdr with an empirical null.
#' @param medv (default 0.5) the cutoff value of explanatory.variable (after 0-1 normalization) for determining if enough pixels for each cell type
#' have explanatory-variable greater than or less than this value (minimum cell_type_threshold/2 required).
#' @return an \code{\linkS4class{RCTD}} object containing the results of the CSIDE algorithm. Contains objects \code{de_results},
#' which contain the results of the CSIDE algorithm including `gene_fits`, which contains the results of fits on individual genes,
#' in addition `sig_gene_list`, a list, for each cell type, of significant genes detected by CSIDE.
#' Additionally, the object contains `internal_vars_de` a list of variables that are used internally by CSIDE
#' @export
run.CSIDE.single <- function(myRCTD, explanatory.variable,  cell_types = NULL, cell_type_threshold = 125,
                          gene_threshold = 5e-5, doublet_mode = T, weight_threshold = NULL,
                          sigma_gene = T, PRECISION.THRESHOLD = 0.05, cell_types_present = NULL, fdr = .01,
                          test_genes_sig = T, normalize_expr = F, logs=F, log_fc_thresh = 0.4, test_error = F, fdr_method = 'BH', medv = 0.5) {
  X2 <- build.designmatrix.single(myRCTD, explanatory.variable)
  barcodes <- rownames(X2)
  explanatory.variable <- explanatory.variable[barcodes]
  region_thresh <- cell_type_threshold / 2
  r1 <- barcodes[X2[,2] < medv]
  if(length(r1) < region_thresh)
    stop(paste0('run.CSIDE.single: number of pixels with explanatory.variable at least medv = ',medv,
                " is less than (one half of cell_type_threshold) = ", region_thresh,
                ". Please make sure that explanatory.variable attains a large value sufficiently often."))
  cell_type_filter <- aggregate_cell_types(myRCTD, r1, doublet_mode = doublet_mode) >= region_thresh
  r2 <- barcodes[X2[,2] > medv]
  if(length(r2) < region_thresh)
    stop(paste0('run.CSIDE.single: number of pixels with explanatory.variable below medv = ',medv,
                " is less than (one half of cell_type_threshold) = ", region_thresh,
                ". Please make sure that explanatory.variable attains a small value sufficiently often."))
  cell_type_filter <- cell_type_filter & (aggregate_cell_types(myRCTD, r2, doublet_mode = doublet_mode) >= region_thresh)
  message(paste0('run.CSIDE.single: filtered out cell types: ', list(which(!cell_type_filter)),
                 ' due to not having sufficiently many pixels with explanatory.value on either side of medv = ',medv,
                 '. Please note that it is required to have on either side of medv at least (one half of cell_type_threshold) = ',
                 region_thresh, ' pixels.'))
  return(run.CSIDE(myRCTD, X2, barcodes, cell_types, gene_threshold = gene_threshold, cell_type_threshold = cell_type_threshold,
                       doublet_mode = doublet_mode, test_mode = 'individual', params_to_test = 2,
                       weight_threshold = weight_threshold, sigma_gene = sigma_gene, test_genes_sig = test_genes_sig,
                       PRECISION.THRESHOLD = PRECISION.THRESHOLD,
                       cell_types_present = cell_types_present, fdr = fdr, normalize_expr = normalize_expr,
                   logs=logs, cell_type_filter = cell_type_filter, log_fc_thresh = log_fc_thresh, test_error = test_error, fdr_method = fdr_method))
}

#' Runs CSIDE on a \code{\linkS4class{RCTD}} object with only an intercept term
#'
#' Identifies cell type specific gene expression for each cell type.
#'
#' The design matrix contains an intercept column only. Uses maximum
#' likelihood estimation to estimate gene expression and standard errors for each gene and each cell type.
#'
#' @param myRCTD an \code{\linkS4class{RCTD}} object with annotated cell types e.g. from the \code{\link{run.RCTD}} function.
#' @param barcodes (default NULL) the barcodes, or pixel names, of the \code{\linkS4class{SpatialRNA}} object to be used when creating the design matrix.
#' @param cell_types the cell types used for CSIDE. If null, cell types will be chosen with aggregate occurrences of
#' at least `cell_type_threshold`, as aggregated by \code{\link{aggregate_cell_types}}
#' @param cell_type_threshold (default 125) min occurrence of number of cells for each cell type to be used, as aggregated by \code{\link{aggregate_cell_types}}
#' @param gene_threshold (default 5e-5) minimum average normalized expression required for selecting genes
#' @param doublet_mode (default TRUE) if TRUE, uses RCTD doublet mode weights. Otherwise, uses RCTD full mode weights
#' @param sigma_gene (default TRUE) if TRUE, fits gene specific overdispersion parameter. If FALSE, overdispersion parameter is same across all genes.
#' @param weight_threshold (default NULL) the threshold of total normalized weights across all cell types
#' in \code{cell_types} per pixel to be included in the model. Default 0.99 for doublet_mode or 0.8 for full_mode.
#' @param PRECISION.THRESHOLD (default 0.05) for checking for convergence, the maximum parameter change per algorithm step
#' @param cell_types_present cell types (a superset of `cell_types`) to be considered as occurring often enough
#' to consider for gene expression contamination during the step filtering out marker genes of other cell types.
#' @param normalize_expr (default FALSE) if TRUE, constrains total gene expression to sum to 1 in each condition.
#' @param logs (default FALSE) if TRUE, writes progress to logs/de_logs.txt
#' @param test_error (default FALSE) if TRUE, exits after testing for error messages without running CSIDE.
#' If set to TRUE, this can be used to quickly evaluate if CSIDE will run without error.
#' @return an \code{\linkS4class{RCTD}} object containing the results of the CSIDE algorithm. Contains objects \code{de_results},
#' which contain the results of the CSIDE algorithm including `gene_fits`, which contains the results of fits on individual genes,
#' in addition `sig_gene_list`, a list, for each cell type, of significant genes detected by CSIDE.
#' Additionally, the object contains `internal_vars_de` a list of variables that are used internally by CSIDE
#' @export
run.CSIDE.intercept <- function(myRCTD, barcodes = NULL, cell_types = NULL, cell_type_threshold = 125,
                             gene_threshold = 5e-5, doublet_mode = T, weight_threshold = NULL,
                             sigma_gene = T, PRECISION.THRESHOLD = 0.05, cell_types_present = NULL,
                             normalize_expr = F, logs=F, test_error = F) {
  X2 <- build.designmatrix.intercept(myRCTD, barcodes = barcodes)
  barcodes <- rownames(X2)
  return(run.CSIDE(myRCTD, X2, barcodes, cell_types, gene_threshold = gene_threshold, cell_type_threshold = cell_type_threshold,
                   doublet_mode = doublet_mode, test_mode = 'individual', params_to_test = 1,
                   weight_threshold = weight_threshold, sigma_gene = sigma_gene, test_genes_sig = FALSE,
                   PRECISION.THRESHOLD = PRECISION.THRESHOLD,
                   cell_types_present = cell_types_present, normalize_expr = normalize_expr,
                   logs=logs, test_error = test_error))
}

#' Runs CSIDE on a \code{\linkS4class{RCTD}} object to detect nonparametric smooth gene expression patterns
#'
#' Identifies cell type specific smooth gene expression patterns. The design matrix contains thin plate spline
#' basis functions spanning the space of smooth functions. Uses maximum likelihood estimation to estimate
#' DE and standard errors for each gene and each cell type. Selects genes with significant nonzero DE.
#'
#' @param myRCTD an \code{\linkS4class{RCTD}} object with annotated cell types e.g. from the \code{\link{run.RCTD}} function.
#' @param df (default 15) the degrees of freedom, or number of basis functions to be used in the model.
#' @param barcodes the barcodes, or pixel names, of the \code{\linkS4class{SpatialRNA}} object to be used when fitting the model.
#' @param cell_types the cell types used for CSIDE. If null, cell types will be chosen with aggregate occurences of
#' at least `cell_type_threshold`, as aggregated by \code{\link{aggregate_cell_types}}
#' @param cell_type_threshold (default 125) min occurence of number of cells for each cell type to be used, as aggregated by \code{\link{aggregate_cell_types}}
#' @param gene_threshold (default 5e-5) minimum average normalized expression required for selecting genes
#' @param doublet_mode (default TRUE) if TRUE, uses RCTD doublet mode weights. Otherwise, uses RCTD full mode weights
#' @param sigma_gene (default TRUE) if TRUE, fits gene specific overdispersion parameter. If FALSE, overdispersion parameter is same across all genes.
#' @param weight_threshold (default NULL) the threshold of total normalized weights across all cell types
#' in \code{cell_types} per pixel to be included in the model. Default 0.99 for doublet_mode or 0.8 for full_mode.
#' @param PRECISION.THRESHOLD (default 0.05) for checking for convergence, the maximum parameter change per algorithm step
#' @param cell_types_present cell types (a superset of `cell_types`) to be considered as occuring often enough
#' to consider for gene expression contamination during the step filtering out marker genes of other cell types.
#' @param fdr (default 0.01) false discovery rate for hypothesis testing
#' @param test_error (default FALSE) if TRUE, exits after testing for error messages without running CSIDE.
#' If set to TRUE, this can be used to quickly evaluate if CSIDE will run without error.
#' @param test_genes_sig (default TRUE) logical controlling whether genes will be tested for significance
#' @return an \code{\linkS4class{RCTD}} object containing the results of the CSIDE algorithm. Contains objects \code{de_results},
#' which contain the results of the CSIDE algorithm including `gene_fits`, which contains the results of fits on individual genes,
#' in addition `sig_gene_list`, a list, for each cell type, of significant genes detected by CSIDE.
#' Additionally, the object contains `internal_vars_de` a list of variables that are used internally by CSIDE
#' @param logs (default FALSE) if TRUE, writes progress to logs/de_logs.txt
#' @export
run.CSIDE.nonparam <- function(myRCTD, df = 15, barcodes = NULL, cell_types = NULL,
                            cell_type_threshold = 125, gene_threshold = 5e-5, doublet_mode = T,
                            weight_threshold = NULL, sigma_gene = T,
                            PRECISION.THRESHOLD = 0.05, cell_types_present = NULL, fdr = .01, test_genes_sig = T,
                            logs=F, test_error = F) {
  X2 <- build.designmatrix.nonparam(myRCTD, barcodes = barcodes, df = df)
  region_thresh <- cell_type_threshold / 4
  barcodes <- rownames(X2)
  coords <- myRCTD@spatialRNA@coords[barcodes,]
  medx <- median(coords$x); medy <- median(coords$y)
  r1 <- barcodes[coords$x < medx & coords$y < medy]
  cell_type_filter <- aggregate_cell_types(myRCTD, r1, doublet_mode = doublet_mode) >= region_thresh
  r2 <- barcodes[coords$x < medx & coords$y > medy]
  cell_type_filter <- cell_type_filter & (aggregate_cell_types(myRCTD, r2, doublet_mode = doublet_mode) >= region_thresh)
  r3 <- barcodes[coords$x > medx & coords$y > medy]
  cell_type_filter <- cell_type_filter & (aggregate_cell_types(myRCTD, r3, doublet_mode = doublet_mode) >= region_thresh)
  r4 <- barcodes[coords$x > medx & coords$y > medy]
  cell_type_filter <- cell_type_filter & (aggregate_cell_types(myRCTD, r4, doublet_mode = doublet_mode) >= region_thresh)
  cell_type_count <- aggregate_cell_types(myRCTD, barcodes, doublet_mode = doublet_mode)
  return(run.CSIDE(myRCTD, X2, barcodes, cell_types, gene_threshold = gene_threshold,
                       doublet_mode = doublet_mode, test_mode = 'individual', cell_type_threshold = cell_type_threshold,
                       weight_threshold = weight_threshold, sigma_gene = sigma_gene,test_genes_sig = test_genes_sig,
                       PRECISION.THRESHOLD = PRECISION.THRESHOLD, test_error = test_error,
                       cell_types_present = cell_types_present, params_to_test = 2:df, fdr = fdr, normalize_expr = F,
                   logs=logs, cell_type_filter = cell_type_filter))
}

#' Runs CSIDE on a \code{\linkS4class{RCTD}} object for DE across multiple discrete regions
#'
#' Identifies cell type specific differential expression (DE) across multiple discrete regions
#' The design matrix contains for each region a column of 0s and 1s representing membership in that region. Uses maximum
#' likelihood estimation to estimate DE and standard errors for each gene and each cell type. Selects
#' genes with significant nonzero DE. Tests for differences in gene expression across regions.
#'
#' @param myRCTD an \code{\linkS4class{RCTD}} object with annotated cell types e.g. from the \code{\link{run.RCTD}} function.
#' @param region_list a list of \code{character} vectors, where each vector contains pixel names, or barcodes, for a single region. These pixel names
#' should be a subset of the pixels in the \code{\linkS4class{SpatialRNA}} object
#' @param cell_types the cell types used for CSIDE. If null, cell types will be chosen with aggregate occurences of
#' at least `cell_type_threshold`, as aggregated by \code{\link{aggregate_cell_types}}
#' @param cell_type_threshold (default 125) min occurence of number of cells for each cell type to be used, as aggregated by \code{\link{aggregate_cell_types}}
#' @param gene_threshold (default 5e-5) minimum average normalized expression required for selecting genes
#' @param doublet_mode (default TRUE) if TRUE, uses RCTD doublet mode weights. Otherwise, uses RCTD full mode weights
#' @param sigma_gene (default TRUE) if TRUE, fits gene specific overdispersion parameter. If FALSE, overdispersion parameter is same across all genes.
#' @param weight_threshold (default NULL) the threshold of total normalized weights across all cell types
#' in \code{cell_types} per pixel to be included in the model. Default 0.99 for doublet_mode or 0.8 for full_mode.
#' @param PRECISION.THRESHOLD (default 0.05) for checking for convergence, the maximum parameter change per algorithm step
#' @param cell_types_present cell types (a superset of `cell_types`) to be considered as occuring often enough
#' to consider for gene expression contamination during the step filtering out marker genes of other cell types.
#' @param fdr (default 0.01) false discovery rate for hypothesis testing
#' @param test_genes_sig (default TRUE) logical controlling whether genes will be tested for significance
#' @param logs (default FALSE) if TRUE, writes progress to logs/de_logs.txt
#' @param test_error (default FALSE) if TRUE, exits after testing for error messages without running CSIDE.
#' If set to TRUE, this can be used to quickly evaluate if CSIDE will run without error.
#' @param log_fc_thresh (default 0.4) the natural log fold change cutoff for differential expression
#' @return an \code{\linkS4class{RCTD}} object containing the results of the CSIDE algorithm. Contains objects \code{de_results},
#' which contain the results of the CSIDE algorithm including `gene_fits`, which contains the results of fits on individual genes,
#' in addition `sig_gene_list`, a list, for each cell type, of significant genes detected by CSIDE.
#' Additionally, the object contains `internal_vars_de` a list of variables that are used internally by CSIDE
#' @export
run.CSIDE.regions <- function(myRCTD, region_list, cell_types = NULL,
                           cell_type_threshold = 125, gene_threshold = 5e-5, doublet_mode = T,
                          weight_threshold = NULL, sigma_gene = T,
                           PRECISION.THRESHOLD = 0.05, cell_types_present = NULL, fdr = 0.01, test_genes_sig = T,
                          logs=F, log_fc_thresh = 0.4, test_error = F) {
  X2 <- build.designmatrix.regions(myRCTD, region_list)
  barcodes <- rownames(X2)
  return(run.CSIDE(myRCTD, X2, barcodes, cell_types, cell_type_threshold = cell_type_threshold, gene_threshold = gene_threshold,
                       doublet_mode = doublet_mode, test_mode = 'categorical',
                       weight_threshold = weight_threshold, sigma_gene = sigma_gene, params_to_test = 1:dim(X2)[2],
                       PRECISION.THRESHOLD = PRECISION.THRESHOLD,test_genes_sig = test_genes_sig,
                       cell_types_present = cell_types_present, fdr = fdr, normalize_expr = F,
                   logs=logs, log_fc_thresh = log_fc_thresh, test_error = test_error))
}

#' Runs cell type specific CSIDE on a \code{\linkS4class{RCTD}} object with a general design matrix
#'
#' Identifies cell type specific differential expression (DE) across a general design matrix of covariates. Uses maximum
#' likelihood estimation to estimate DE and standard errors for each gene and each cell type. Selects
#' genes with significant nonzero DE. The type of test is determined by \code{test_mode}, and the parameters tested
#' is determined by \code{params_to_test}.
#'
#' @param myRCTD an \code{\linkS4class{RCTD}} object with annotated cell types e.g. from the \code{\link{run.RCTD}} function.
#' @param X a matrix containing the covariates for running CSIDE. The rownames represent pixel names and
#' should be a subset of the pixels in the \code{\linkS4class{SpatialRNA}} object. The columns each represent a covariate for
#' explaining differential expression and need to be linearly independent.
#' @param barcodes the barcodes, or pixel names, of the \code{\linkS4class{SpatialRNA}} object to be used when fitting the model.
#' @param cell_types the cell types used for CSIDE. If null, cell types will be chosen with aggregate occurences of
#' at least `cell_type_threshold`, as aggregated by \code{\link{aggregate_cell_types}}
#' @param cell_type_specific: (default TRUE for all covariates). A logical vector of length the number of covariates
#' indicating whether each covariate's DE parameters should be cell type-specific or shared across all cell types.
#' @param params_to_test: (default 2 for test_mode = 'individual', all parameters for test_mode = 'categorical'). An integer vector of parameter
#' indices to test. For example c(1,4,5) would test only parameters corresponding to columns 1, 4, and 5 of the design matrix.
#' @param cell_type_threshold (default 125) min occurence of number of cells for each cell type to be used, as aggregated by \code{\link{aggregate_cell_types}}
#' @param gene_threshold (default 5e-5) minimum average normalized expression required for selecting genes
#' @param doublet_mode (default TRUE) if TRUE, uses RCTD doublet mode weights. Otherwise, uses RCTD full mode weights
#' @param sigma_gene (default TRUE) if TRUE, fits gene specific overdispersion parameter. If FALSE, overdispersion parameter is same across all genes.
#' @param weight_threshold (default NULL) the threshold of total normalized weights across all cell types
#' in \code{cell_types} per pixel to be included in the model. Default 0.99 for doublet_mode or 0.8 for full_mode.
#' @param test_mode (default 'individual') if 'individual', tests for DE individually for each parameter. If 'categorical', then tests for differences
#' across multiple categorical parameters
#' @param PRECISION.THRESHOLD (default 0.05) for checking for convergence, the maximum parameter change per algorithm step
#' @param cell_types_present cell types (a superset of `cell_types`) to be considered as occuring often enough
#' to consider for gene expression contamination during the step filtering out marker genes of other cell types.
#' @param fdr (default 0.01) false discovery rate for hypothesis testing
#' @param normalize_expr (default FALSE) if TRUE, constrains total gene expression to sum to 1 in each condition.
#' Setting normalize_expr = TRUE is only valid for testing single parameters with test_mode = 'individual'.
#' @param test_genes_sig (default TRUE) logical controlling whether genes will be tested for significance
#' @param logs (default FALSE) if TRUE, writes progress to logs/de_logs.txt
#' @param test_error (default FALSE) if TRUE, exits after testing for error messages without running CSIDE.
#' If set to TRUE, this can be used to quickly evaluate if CSIDE will run without error.
#' @param log_fc_thresh (default 0.4) the natural log fold change cutoff for differential expression
#' @param fdr_method (default BH) if BH, uses the Benjamini-Hochberg method. Otherwise, uses local fdr with an empirical null.
#' @return an \code{\linkS4class{RCTD}} object containing the results of the CSIDE algorithm. Contains objects \code{de_results},
#' which contain the results of the CSIDE algorithm including `gene_fits`, which contains the results of fits on individual genes,
#' in addition `sig_gene_list`, a list, for each cell type, of significant genes detected by CSIDE.
#' Additionally, the object contains `internal_vars_de` a list of variables that are used internally by CSIDE
#' @export
run.CSIDE <- function(myRCTD, X, barcodes, cell_types = NULL, gene_threshold = 5e-5, cell_type_threshold = 125,
                          doublet_mode = T, test_mode = 'individual', weight_threshold = NULL,
                          sigma_gene = T, PRECISION.THRESHOLD = 0.05, cell_types_present = NULL,
                          test_genes_sig = T, fdr = .01, cell_type_specific = NULL,
                      params_to_test = NULL, normalize_expr = F, logs=F, log_fc_thresh = 0.4,
                      cell_type_filter = NULL, test_error = F, fdr_method = 'BH') {
  X <- check_designmatrix(X, 'run.CSIDE', require_2d = TRUE)
  if(is.null(cell_type_specific))
    cell_type_specific <- !logical(dim(X)[2])
  check_cell_type_specific(cell_type_specific, dim(X)[2])
  X1 <- X[,!cell_type_specific];
  if(any(!cell_type_specific))
    X2 <- X[,cell_type_specific]
  else
    X2 <- X
  return(run.CSIDE.general(myRCTD, X1, X2, barcodes, cell_types, cell_type_threshold = cell_type_threshold,
                           gene_threshold = gene_threshold,
                       doublet_mode = doublet_mode, test_mode = test_mode, weight_threshold = weight_threshold,
                       sigma_gene = sigma_gene, PRECISION.THRESHOLD = PRECISION.THRESHOLD, params_to_test = params_to_test,
                       cell_types_present = cell_types_present, test_genes_sig = test_genes_sig,
                       fdr = fdr, normalize_expr = normalize_expr, logs=logs,
                       cell_type_filter = cell_type_filter, log_fc_thresh = log_fc_thresh, test_error = test_error, fdr_method = fdr_method))
}

#' Runs CSIDE on a \code{\linkS4class{RCTD}} object with a general design matrix
#'
#' Identifies differential expression (DE) across a general design matrix of covariates. DE parameters can be
#' cell type-specific or shared across all cell types. Uses maximum
#' likelihood estimation to estimate DE and standard errors for each gene and each cell type. Selects
#' genes with significant nonzero DE. The type of test is determined by \code{test_mode}, and the parameters tested
#' is determined by \code{params_to_test}.
#'
#' @param myRCTD an \code{\linkS4class{RCTD}} object with annotated cell types e.g. from the \code{\link{run.RCTD}} function.
#' @param X1 a matrix containing the covariates shared across all cell types. The rownames represent pixel names and
#' should be a subset of the pixels in the \code{\linkS4class{SpatialRNA}} object. The columns each represent a covariate for
#' explaining differential expression and need to be linearly independent.
#' @param X2 a matrix containing the cell type-specific covariates. The rownames represent pixel names and
#' should be a subset of the pixels in the \code{\linkS4class{SpatialRNA}} object. The columns each represent a covariate for
#' explaining differential expression and need to be linearly independent.
#' @param barcodes the barcodes, or pixel names, of the \code{\linkS4class{SpatialRNA}} object to be used when fitting the model.
#' @param cell_types the cell types used for CSIDE. If null, cell types will be chosen with aggregate occurences of
#' at least `cell_type_threshold`, as aggregated by \code{\link{aggregate_cell_types}}
#' @param params_to_test: (default 2 for test_mode = 'individual', all parameters for test_mode = 'categorical'). An integer vector of parameter
#' indices to test. For example c(1,4,5) would test only parameters corresponding to columns 1, 4, and 5 of the design matrix X2.
#' @param cell_type_threshold (default 125) min occurence of number of cells for each cell type to be used, as aggregated by \code{\link{aggregate_cell_types}}
#' @param gene_threshold (default 5e-5) minimum average normalized expression required for selecting genes
#' @param doublet_mode (default TRUE) if TRUE, uses RCTD doublet mode weights. Otherwise, uses RCTD full mode weights
#' @param sigma_gene (default TRUE) if TRUE, fits gene specific overdispersion parameter. If FALSE, overdispersion parameter is same across all genes.
#' @param weight_threshold (default NULL) the threshold of total normalized weights across all cell types
#' in \code{cell_types} per pixel to be included in the model. Default 0.99 for doublet_mode or 0.8 for full_mode.
#' @param test_mode (default 'individual') if 'individual', tests for DE individually for each parameter. If 'categorical', then tests for differences
#' across multiple categorical parameters
#' @param PRECISION.THRESHOLD (default 0.05) for checking for convergence, the maximum parameter change per algorithm step
#' @param cell_types_present cell types (a superset of `cell_types`) to be considered as occuring often enough
#' to consider for gene expression contamination during the step filtering out marker genes of other cell types.
#' @param fdr (default 0.01) false discovery rate for hypothesis testing
#' @param test_genes_sig (default TRUE) logical controlling whether genes will be tested for significance
#' @param normalize_expr (default FALSE) if TRUE, constrains total gene expression to sum to 1 in each condition.
#' Setting normalize_expr = TRUE is only valid for testing single parameters with test_mode = 'individual'.
#' @param logs (default FALSE) if TRUE, writes progress to logs/de_logs.txt
#' @param log_fc_thresh (default 0.4) the natural log fold change cutoff for differential expression
#' @param test_error (default FALSE) if TRUE, exits after testing for error messages without running CSIDE.
#' If set to TRUE, this can be used to quickly evaluate if CSIDE will run without error.
#' @param fdr_method (default BH) if BH, uses the Benjamini-Hochberg method. Otherwise, uses local fdr with an empirical null.
#' @return an \code{\linkS4class{RCTD}} object containing the results of the CSIDE algorithm. Contains objects \code{de_results},
#' which contain the results of the CSIDE algorithm including `gene_fits`, which contains the results of fits on individual genes,
#' in addition `sig_gene_list`, a list, for each cell type, of significant genes detected by CSIDE, whereas
#' `all_gene_list` is the analogous list for all genes (including nonsignificant).
#' Additionally, the object contains `internal_vars_de` a list of variables that are used internally by CSIDE
#' @export
run.CSIDE.general <- function(myRCTD, X1, X2, barcodes, cell_types = NULL, gene_threshold = 5e-5, cell_type_threshold = 125,
                          doublet_mode = T, test_mode = 'individual', weight_threshold = NULL,
                          sigma_gene = T, PRECISION.THRESHOLD = 0.05, cell_types_present = NULL,
                          test_genes_sig = T, fdr = .01, params_to_test = NULL, normalize_expr = F,
                          logs=F, cell_type_filter = NULL, log_fc_thresh = 0.4, test_error = FALSE, fdr_method = 'BH') {
  if(gene_threshold == .01 || fdr == 0.25 || cell_type_threshold <= 10 ||
     (!is.null(weight_threshold) && weight_threshold == 0.1))
    warning('run.CSIDE.general: some parameters are set to the CSIDE vignette values, which are intended for testing but not proper execution. For more accurate results, consider using the default parameters to this function.')
  else if(weight_threshold < 0.75)
    warning('run.CSIDE.general: we recommend setting weight_threshold to at least 0.75 since otherwise cell types not included in the model will have large proportions.')
  if(doublet_mode && myRCTD@config$RCTDmode != 'doublet')
    stop('run.CSIDE.general: attempted to run CSIDE in doublet mode, but RCTD was not run in doublet mode. Please run CSIDE in full mode (doublet_mode = F) or run RCTD in doublet mode.')
  if(!any("cell_types_assigned" %in% names(myRCTD@internal_vars)) || !myRCTD@internal_vars$cell_types_assigned)
    stop('run.CSIDE.general: cannot run CSIDE unless cell types have been assigned. If cell types have been assigned, you may run "myRCTD <- set_cell_types_assigned(myRCTD)".')
  if((myRCTD@config$RCTDmode != 'multi') && (length(setdiff(barcodes,rownames(myRCTD@results$weights))) > 0)) {
    warning('run.CSIDE.general: some elements of barcodes do not appear in myRCTD object (myRCTD@results$weights), but they are required to be a subset. Downsampling barcodes to the intersection of the two sets.')
    barcodes <- intersect(barcodes,rownames(myRCTD@results$weights))
  }
  cell_type_info <- myRCTD@cell_type_info$info
  if(doublet_mode) {
    my_beta <- get_beta_doublet(barcodes, cell_type_info[[2]], myRCTD@results$results_df, myRCTD@results$weights_doublet)
    thresh <- 0.999
  } else if(myRCTD@config$RCTDmode == "multi") {
    my_beta <- get_beta_multi(barcodes, cell_type_info[[2]], myRCTD@results, myRCTD@spatialRNA@coords)
    thresh <- 0.999
  } else {
    my_beta <- as.matrix(sweep(myRCTD@results$weights, 1, rowSums(myRCTD@results$weights), '/'))
    thresh <- 0.8
  }
  if(!is.null(weight_threshold))
    thresh <- weight_threshold
  cell_types <- choose_cell_types(myRCTD, barcodes, doublet_mode, cell_type_threshold, cell_types,
                                  my_beta, thresh, cell_type_filter)
  if(length(cell_types) < 1)
    stop('run.CSIDE.general: zero cell types remain. Cannot run CSIDE with zero cell types.')
  message(paste0("run.CSIDE.general: running CSIDE with cell types ",paste(cell_types, collapse = ', ')))
  X1 <- check_designmatrix(X1, 'run.CSIDE.general')
  X2 <- check_designmatrix(X2, 'run.CSIDE.general', require_2d = TRUE)
  if(!(test_mode %in% c('individual', 'categorical')))
    stop(c('run.CSIDE.general: not valid test_mode = ',test_mode,'. Please set test_mode = "categorical" or "individual".'))
  if(is.null(params_to_test))
    if(test_mode == 'individual')
      params_to_test <- min(2, dim(X2)[2])
    else
      params_to_test <- 1:dim(X2)[2]
  if(normalize_expr && (test_mode != 'individual' || length(params_to_test) > 1))
    stop('run.CSIDE.general: Setting normalize_expr = TRUE is only valid for testing single parameters with test_mode = individual')
  message(paste0("run.CSIDE.general: configure params_to_test = ",
                 paste(paste0(params_to_test, ', ', collapse = ""))))
  if(any(!(params_to_test %in% 1:dim(X2)[2])))
    stop(c('run.CSIDE.general: params_to_test must be a vector of integers from 1 to dim(X2)[2] = ', dim(X2)[2],
           'please make sure that tested parameters are in the required range.'))
  if(test_mode == 'categorical' && any(!(X2[,params_to_test] %in% c(0,1))))
    stop(c('run.CSIDE.general: for test_mode = categorical, colums params_to_test, ',params_to_test,', must have values 0 or 1.'))
  if(is.null(cell_types_present))
    cell_types_present <- cell_types
  if(any(!(barcodes %in% rownames(X1))) || any(!(barcodes %in% rownames(X2))))
    stop('run.CSIDE.general: some barcodes do not appear in the rownames of X1 or X2.')
  puck = myRCTD@originalSpatialRNA
  gene_list_tot <- filter_genes(puck, threshold = gene_threshold)
  if(length(gene_list_tot) == 0)
    stop('run.CSIDE.general: no genes past threshold. Please consider lowering gene_threshold.')
  if(length(intersect(gene_list_tot,rownames(cell_type_info[[1]]))) == 0)
    stop('run.CSIDE.general: no genes that past threshold were contained in the single cell reference. Please lower gene threshold or ensure that there is agreement between the single cell reference genes and the SpatialRNA genes.')
  nUMI <- puck@nUMI[barcodes]
  res <- filter_barcodes_cell_types(barcodes, cell_types, my_beta, thresh = thresh)
  if(test_error)
    return(myRCTD)
  barcodes <- res$barcodes; my_beta <- res$my_beta
  sigma_init <- as.character(100*myRCTD@internal_vars$sigma)
  if(sigma_gene) {
    set_global_Q_all()
    sigma_set <- sigma_init
    set_likelihood_vars(Q_mat_all[[sigma_init]], X_vals, sigma = sigma_set)
  } else {
    set_likelihood_vars_sigma(sigma_init)
  }
  gene_fits <- get_de_gene_fits(X1[barcodes, , drop = FALSE],X2[barcodes, , drop = FALSE],my_beta, nUMI[barcodes], gene_list_tot,
                                cell_types, restrict_puck(puck, barcodes), barcodes, sigma_init,
                                test_mode, numCores = myRCTD@config$max_cores, sigma_gene = sigma_gene,
                                PRECISION.THRESHOLD = PRECISION.THRESHOLD, params_to_test = params_to_test,
                                logs=logs)
  if(normalize_expr)
    myRCTD <- normalize_de_estimates(myRCTD, normalize_expr = normalize_expr,
                                     param_position = params_to_test)
  if(test_genes_sig) {
    both_gene_list <- get_sig_genes(puck, myRCTD, gene_list_tot, cell_types, my_beta, barcodes, nUMI,
                            gene_fits, cell_types_present, X2, test_mode, fdr = fdr,
                            params_to_test = params_to_test, normalize_expr = normalize_expr,
                            log_fc_thresh = log_fc_thresh, fdr_method = fdr_method)
    sig_gene_list <- both_gene_list$sig_gene_list; all_gene_list <- both_gene_list$all_gene_list
  } else {
    sig_gene_list <- NULL
    all_gene_list <- NULL
  }
  myRCTD@internal_vars_de <- list(barcodes = barcodes, cell_types = cell_types, doublet_mode = doublet_mode,
                                  cell_types_present = cell_types_present,
                                  my_beta = my_beta, X1 = X1, X2 = X2,
                                  test_mode = test_mode, params_to_test = params_to_test)
  myRCTD@de_results <- list(gene_fits = gene_fits, sig_gene_list = sig_gene_list, all_gene_list = all_gene_list)
  return(myRCTD)
}

get_sig_genes <- function(puck, myRCTD, gene_list_tot, cell_types, my_beta, barcodes, nUMI,
                          gene_fits, cell_types_present, X2, test_mode, params_to_test = 2,
                          fdr = .01, p_thresh = 1, log_fc_thresh = 0.4, normalize_expr = F, fdr_method = 'BH') {
  cti_renorm <- get_norm_ref(puck, myRCTD@cell_type_info$info[[1]], intersect(gene_list_tot,rownames(myRCTD@cell_type_info$info[[1]])), myRCTD@internal_vars$proportions)
  sig_gene_list <- list(); all_gene_list <- list()
  for(cell_type in cell_types) {
    gene_list_type <- get_gene_list_type(my_beta, barcodes, cell_type, nUMI, gene_list_tot,
                                         cti_renorm, cell_types_present, gene_fits, test_mode = test_mode)
    if(test_mode == 'individual')
      both_genes <- find_sig_genes_individual(cell_type, cell_types, gene_fits, gene_list_type, X2,
                                         params_to_test = params_to_test, fdr = fdr, p_thresh = p_thresh,
                                         log_fc_thresh = log_fc_thresh, normalize_expr = normalize_expr,
                                         fdr_method = fdr_method)
    else if(test_mode == 'categorical') {
      both_genes <- find_sig_genes_categorical(cell_type, cell_types, gene_fits, gene_list_type, X2,
                                        p_thresh = p_thresh, log_fc_thresh = log_fc_thresh,
                                        params_to_test = params_to_test)
    }
    sig_genes <- both_genes$sig_genes; all_genes <- both_genes$all_genes
    sig_gene_list[[cell_type]] <- sig_genes
    all_gene_list[[cell_type]] <- all_genes
  }
  return(list(sig_gene_list = sig_gene_list, all_gene_list = all_gene_list))
}

test_genes_sig_post <- function(myRCTD, params_to_test = NULL, fdr = .01, p_thresh = 1,
                                log_fc_thresh = 0.4, normalize_expr = F, fdr_method = 'BH') {
  puck <- myRCTD@originalSpatialRNA
  gene_list_tot <- rownames(myRCTD@de_results$gene_fits$s_mat)
  cell_types <- myRCTD@internal_vars_de$cell_types
  my_beta <- myRCTD@internal_vars_de$my_beta
  X2 <- myRCTD@internal_vars_de$X2
  barcodes <- myRCTD@internal_vars_de$barcodes
  nUMI <- puck@nUMI[barcodes]
  test_mode <- myRCTD@internal_vars_de$test_mode
  cell_types_present <- myRCTD@internal_vars_de$cell_types_present
  gene_fits <- myRCTD@de_results$gene_fits
  if(is.null(params_to_test))
    params_to_test <- myRCTD@internal_vars_de$params_to_test
  both_gene_list <- get_sig_genes(puck, myRCTD, gene_list_tot, cell_types, my_beta, barcodes, nUMI,
                                  gene_fits, cell_types_present, X2, test_mode,
                                  params_to_test = params_to_test, fdr = fdr,
                                 p_thresh = p_thresh, log_fc_thresh = log_fc_thresh,
                                 normalize_expr = normalize_expr, fdr_method = fdr_method)
  myRCTD@de_results$sig_gene_list <- both_gene_list$sig_gene_list
  myRCTD@de_results$all_gene_list <- both_gene_list$all_gene_list
  return(myRCTD)
}

find_sig_genes_categorical <- function(cell_type, cell_types, gene_fits, gene_list_type, X2, fdr = 0.01,
                                p_thresh = 1, log_fc_thresh = 0.4, params_to_test = NULL) {
  if(length(gene_list_type) == 0)
    stop(paste0('find_sig_genes_categorical: cell type ', cell_type,
                ' has not converged on any genes. Consider removing this cell type from the model using the cell_types option.'))
  if(is.null(params_to_test))
    params_to_test <- 1:dim(X2)[2]
  n_regions <- length(params_to_test); n_cell_types <- length(cell_types)
  cell_ind = (which(cell_types == cell_type))
  s_mat_ind <- (1:dim(X2)[2]) + (n_regions*(cell_ind - 1))
  p_val_sig_pair <- numeric(length(gene_list_type)); names(p_val_sig_pair) <- gene_list_type
  log_fc_best_pair <- numeric(length(gene_list_type)); names(log_fc_best_pair) <- gene_list_type
  sd_vec <- numeric(length(gene_list_type)); names(sd_vec) <- gene_list_type
  sd_lfc_vec <- numeric(length(gene_list_type)); names(sd_lfc_vec) <- gene_list_type
  i1_vec <- numeric(length(gene_list_type)); names(i1_vec) <- gene_list_type
  i2_vec <- numeric(length(gene_list_type)); names(i2_vec) <- gene_list_type
  for(gene in gene_list_type) {
    con_regions <- get_con_regions(gene_fits, gene, dim(X2)[2], cell_ind, n_cell_types) &
      (params_to_test %in% 1:dim(X2)[2])
    n_regions_con <- sum(con_regions)
    x <- gene_fits$all_vals[gene, con_regions,cell_ind]
    s_mat_ind_cur <- s_mat_ind[con_regions]
    var_vals <- (gene_fits$s_mat[gene, s_mat_ind_cur])^2
    ovr_best_p_val <- 1
    best_log_fc <- 0; best_sd <- 0
    best_i1 <- 0; best_i2 <- 0
    for(i1 in 1:(n_regions_con-1))
      for(i2 in (i1+1):n_regions_con) {
        log_fc <- abs(x[i1] - x[i2])
        sd_cur <- sqrt(var_vals[i1] + var_vals[i2])
        z_score <- (log_fc) / sd_cur
        p_val <- 2*(pnorm(-z_score))
        if(p_val < ovr_best_p_val) {
          ovr_best_p_val <- p_val
          best_log_fc <- log_fc
          best_sd <- sd_cur
          best_i1 <- which(con_regions)[i1]
          best_i2 <- which(con_regions)[i2]
        }
      }
    p_val_sig_pair[gene] <- min(1, ovr_best_p_val * choose(n_regions_con, 2))
    log_fc_best_pair[gene] <- best_log_fc
    sd_vec[gene] <- best_sd
    sd_lfc_vec[gene] <- sd(x)
    i1_vec[gene] <- best_i1
    i2_vec[gene] <- best_i2
  }
  gene_list_sig <- fdr_sig_genes(gene_list_type, p_val_sig_pair, fdr)
  all_genes <- data.frame(sd_lfc_vec[gene_list_type], i1_vec[gene_list_type], i2_vec[gene_list_type],
                          sd_vec[gene_list_type], p_val_sig_pair[gene_list_type],
                          log_fc_best_pair[gene_list_type])
  rownames(all_genes) <- gene_list_type
  custom_names <- c('sd_lfc','paramindex1_best', 'paramindex2_best', 'sd_best','p_val_best','log_fc_best')
  colnames(all_genes) <- custom_names
  if(length(gene_list_type) > 1) {
    all_genes <- data.frame(all_genes, gene_fits$all_vals[rownames(all_genes),params_to_test,cell_ind],
                            gene_fits$s_mat[rownames(all_genes),s_mat_ind[params_to_test]]) # add on the means
    colnames(all_genes)[(length(custom_names)+1):length(all_genes)] <-
      c(lapply(params_to_test,function(x) paste0('mean_',x)), lapply(params_to_test,function(x) paste0('sd_',x)))
  } else {
    if(length(gene_list_type) == 1) {
      all_genes <- data.frame(t(unlist((c(all_genes, gene_fits$all_vals[rownames(all_genes),params_to_test,cell_ind],
                                          gene_fits$s_mat[rownames(all_genes),s_mat_ind[params_to_test]])))))
      rownames(all_genes) <- gene_list_type
      colnames(all_genes)[(length(custom_names)+1):length(all_genes)] <-
        c(lapply(params_to_test,function(x) paste0('mean_',x)), lapply(params_to_test,function(x) paste0('sd_',x)))
    } else
      all_genes <- list()
  }
  if(length(gene_list_sig) > 0) {
    sig_genes <- all_genes[gene_list_sig, ]
    sig_genes <- sig_genes[abs(sig_genes$p_val < p_thresh) & abs(sig_genes$log_fc) >= log_fc_thresh, ]
  } else {
    sig_genes <- list()
  }
  return(list(sig_genes = sig_genes, all_genes = all_genes))
}

find_sig_genes_individual <- function(cell_type, cell_types, gene_fits, gene_list_type, X2, params_to_test = 2, fdr = 0.01, p_thresh = 1,
                                      log_fc_thresh = 0.4, normalize_expr = F, fdr_method = 'BH') {
  if(length(gene_list_type) == 0)
    stop(paste0('find_sig_genes_individual: cell type ', cell_type,
                ' has not converged on any genes. Consider removing this cell type from the model using the cell_types option.'))
  ct_ind <- which(cell_types == cell_type)
  I_ind = dim(X2)[2]*(ct_ind - 1) + params_to_test
  I_ind_intercept = dim(X2)[2]*(ct_ind - 1) + 1
  if(normalize_expr) {
    log_fc <- gene_fits$mean_val_cor[[cell_type]][gene_list_type]
  } else {
    log_fc <- gene_fits$all_vals[gene_list_type,params_to_test, ct_ind]
  }
  s_vec <- gene_fits$s_mat[gene_list_type,I_ind]
  z_score <- abs(log_fc) / s_vec
  p_val <- 2*(pnorm(-z_score))
  if(length(params_to_test) > 1)
    p_val <- pmin(apply(p_val, 1, min)*length(params_to_test),1)
  names(p_val) <- gene_list_type
  gene_list_sig <- fdr_sig_genes(gene_list_type, p_val, fdr, Z = log_fc / s_vec, method = fdr_method)
  if(length(gene_list_sig) > 0)
    p_thresh <- min(p_thresh, max(p_val[gene_list_sig]))
  if(length(params_to_test) > 1) {
    p_val <- 2*(pnorm(-z_score))
    best_mat <- function(gene) {
      index <- which(p_val[gene,]*length(params_to_test) < p_thresh)
      if(length(index) > 0) {
        best_ind <- which.max(abs(z_score[gene,index]))
        lfc <- log_fc[gene,index][best_ind]
        sd <- s_vec[gene,index][best_ind]
        z <- z_score[gene,index][best_ind]
        best_ind <- params_to_test[index[best_ind]]
        return(c(best_ind, lfc,sd,z))
      } else {
        return(c(0,0,0,0))
      }
    }
    best_ind <- function(gene) {
      best_mat(gene)[1]
    }
    best_log_fc <- function(gene) {
      best_mat(gene)[2]
    }
    best_sd <- function(gene) {
      best_mat(gene)[3]
    }
    best_Z <- function(gene) {
      best_mat(gene)[4]
    }
    best_indn <- unlist(lapply(gene_list_type, best_ind))
    names(best_indn) <- gene_list_type
    log_fcn <- unlist(lapply(gene_list_type, best_log_fc))
    names(log_fcn) <- gene_list_type
    z_scoren <- unlist(lapply(gene_list_type, best_Z))
    names(z_scoren) <- gene_list_type
    s_vec <- unlist(lapply(gene_list_type, best_sd))
    names(s_vec) <- gene_list_type
    best_ind <- best_indn; z_score <- z_scoren; log_fc <- log_fcn
    p_val <- pmin(apply(p_val, 1, min)*length(params_to_test),1)
  } else {
    best_ind <- rep(params_to_test, length(gene_list_type))
    names(best_ind) <- gene_list_type
  }
  all_genes <- data.frame(z_score[gene_list_type], log_fc[gene_list_type], s_vec[gene_list_type], best_ind[gene_list_type])
  names(all_genes) <- c('Z_score','log_fc', 'se', 'paramindex_best')
  all_genes$conv <- gene_fits$con_mat[gene_list_type, cell_type]
  all_genes$p_val <- p_val[gene_list_type]
  if(length(params_to_test) == 1 & !any(X2[,1] != 1)) {
    mean_0 <- gene_fits$all_vals[gene_list_type, 1, ct_ind]
    mean_1 <- mean_0 + log_fc
    sd_0 <- gene_fits$s_mat[gene_list_type,I_ind_intercept]
    sd_1 <- s_vec^2 - sd_0^2
    sd_1[sd_1 < 0] <- 100
    sd_1 <- sqrt(sd_1)
    all_genes$mean_0 <- mean_0
    all_genes$mean_1 <- mean_1
    all_genes$sd_0 <- sd_0
    all_genes$sd_1 <- sd_1
  }
  sig_genes <- all_genes[gene_list_sig, ]
  return(list(sig_genes = sig_genes, all_genes = all_genes))
}

get_de_gene_fits <- function(X1,X2,my_beta, nUMI, gene_list, cell_types, puck, barcodes, sigma_init, test_mode,
                             numCores = 4, sigma_gene = T, PRECISION.THRESHOLD = 0.05, params_to_test = 2, logs=F) {
  results_list <- fit_de_genes(X1,X2,my_beta, nUMI, gene_list, puck, barcodes,
                               sigma_init, test_mode, numCores = numCores,
                               sigma_gene = sigma_gene,
                               PRECISION.THRESHOLD = PRECISION.THRESHOLD, logs = logs)
  N_genes <- length(results_list)
  intercept_val <- matrix(0,nrow = N_genes, ncol = length(cell_types))
  mean_val <- matrix(0,nrow = N_genes, ncol = length(cell_types))
  all_vals <- array(0, dim = c(N_genes, dim(X2)[2],length(cell_types)))
  dimnames(all_vals)[[1]] <- gene_list
  dimnames(all_vals)[[3]] <- cell_types
  con_val <- logical(N_genes)
  ll_val <- numeric(N_genes)
  n_val <- numeric(N_genes)
  sigma_g <- numeric(N_genes)
  names(sigma_g) <- gene_list
  I_val <- list()
  names(n_val) <- gene_list
  names(con_val) <- gene_list
  names(ll_val) <- gene_list
  rownames(mean_val) <- gene_list; colnames(mean_val) <- cell_types
  rownames(intercept_val) <- gene_list; colnames(intercept_val) <- cell_types
  d_vals <- matrix(0,nrow=N_genes,ncol=dim(X2)[2]*length(cell_types))
  s_mat <- matrix(0, nrow = N_genes, ncol = dim(X2)[2]*length(cell_types))
  precision_mat <- matrix(0, nrow = N_genes, ncol = dim(X2)[2]*length(cell_types))
  con_all <- matrix(FALSE, nrow = N_genes, ncol = dim(X2)[2]*length(cell_types))
  con_mat <- matrix(FALSE, nrow = N_genes, ncol = length(cell_types))
  error_mat <- matrix(FALSE, nrow = N_genes, ncol = length(cell_types))
  rownames(precision_mat) <- gene_list; rownames(con_all) <- gene_list
  rownames(s_mat) <- gene_list; rownames(con_mat) <- gene_list; rownames(error_mat) <- gene_list
  colnames(s_mat) <- get_param_names(X1,X2, cell_types)
  colnames(precision_mat) <- get_param_names(X1,X2, cell_types)
  colnames(con_all) <- get_param_names(X1,X2, cell_types)
  colnames(con_mat) <- cell_types
  colnames(error_mat) <- cell_types
  rownames(d_vals) <- gene_list
  for(i in 1:N_genes) {
    sigma_g[i] <- results_list[[i]]$sigma_s_best
    res <- results_list[[i]]$res
    d_vals[i,] <- res$d
    mean_val[i,] <- res$alpha2[params_to_test[1],]
    intercept_val[i,] <- res$alpha2[1,]
    all_vals[i, ,] <- res$alpha2
    con_val[i] <- res$converged
    precision_mat[i,] <- res$precision
    ll_val[i] <- res$log_l
    n_val[i] <- res$n.iter
    I_val[[i]] <- res$I
    s_mat[i,] <- sqrt(diag(I_val[[i]]))
    con_mat[i,] <- res$converged_vec
    con_all[i,] <- res$precision < PRECISION.THRESHOLD
    error_mat[i,] <- res$error_vec
  }
  return(list(mean_val = mean_val, con_val = con_val, ll_val = ll_val, I_val = I_val, s_mat = s_mat,
              n.iter = n_val,d_vals = d_vals, intercept_val = intercept_val, all_vals = all_vals,
              precision_mat = precision_mat, sigma_g = sigma_g, con_mat = con_mat, con_all = con_all, error_mat = error_mat))
}

fit_de_genes <- function(X1,X2,my_beta, nUMI, gene_list, puck, barcodes, sigma_init, test_mode, numCores = 4, sigma_gene = T, PRECISION.THRESHOLD = 0.05,
                         logs=F) {
  results_list <- list()
  if(numCores == 1) {
    for(i in 1:length(gene_list)) {
      message(i)
      gene <- gene_list[i]
      Y <- puck@counts[gene, barcodes]
      results_list[[i]] <- estimate_gene_wrapper(Y,X1,X2,my_beta, nUMI, sigma_init, test_mode, verbose = F, n.iter = 200, MIN_CHANGE = 1e-3, sigma_gene = sigma_gene, PRECISION.THRESHOLD = PRECISION.THRESHOLD)
    }
  } else {
    cl <- parallel::makeCluster(numCores,setup_strategy = "sequential",outfile="") #makeForkCluster
    doParallel::registerDoParallel(cl)
    environ = c('estimate_effects_trust', 'solveIRWLS.effects_trust', 'K_val','X_vals',
                'calc_log_l_vec', 'get_d1_d2', 'calc_Q_all','psd','construct_hess_fast',
                'choose_sigma_gene', 'estimate_gene_wrapper', 'check_converged_vec', 'calc_log_l_vec_fast')
    if(sigma_gene)
      environ <- c(environ, 'Q_mat_all', 'SQ_mat_all')
    else
      environ <- c(environ, 'Q_mat', 'SQ_mat')
    if (logs) {
      out_file = "logs/de_log.txt"
      if(!dir.exists('logs'))
        dir.create('logs')
      if(file.exists(out_file))
        file.remove(out_file)
    }
    results_list <- foreach::foreach(i = 1:length(gene_list), .packages = c("quadprog", "spacexr", "Rfast"), .export = environ) %dopar% {
      if (logs) {
        if(i %% 1 == 0) { ##10
          cat(paste0("Testing sample: ",i," gene ", gene_list[i],"\n"), file=out_file, append=TRUE)
        }
      }
      assign("X_vals",X_vals, envir = globalenv()); assign("K_val",K_val, envir = globalenv());
      if(sigma_gene) {
        assign("Q_mat_all",Q_mat_all, envir = globalenv());
        assign("SQ_mat_all",SQ_mat_all, envir = globalenv());
      } else {
        assign("Q_mat",Q_mat, envir = globalenv()); assign("SQ_mat",SQ_mat, envir = globalenv())
      }
      gene <- gene_list[i]
      Y <- puck@counts[gene, barcodes]
      res <- estimate_gene_wrapper(Y,X1,X2,my_beta, nUMI, sigma_init, test_mode, verbose = F, n.iter = 200, MIN_CHANGE = 1e-3, sigma_gene = sigma_gene)
    }
    parallel::stopCluster(cl)
  }
  return(results_list)
}