Molecular marker identification and model construction based on gene pair relative Expression order (REOs).
It is based on julia language development, including molecular marker identification and model construction.
Molecular marker gene pairs and model construction are obtained by this function:
RankMS_main(fn_expr::AbstractString = "matrix.mtx", rn_expr::AbstractString = "features.tsv", cn_expr::AbstractString = "barcodes.tsv", fn_meta::AbstractString = "fn_meta.txt", fn_feature::AbstractString = "fn_feature.txt"; building_random_forests::AbstractString = "yes", n_top::Int = 10, n_train::Int = 13, n_test::Int = 3, t_hill_iter_num::Int = 500, t_train_iter_num::Int = 15, ncell_pseudo::Int = 0, fn_meta_delim::AbstractChar = '\t', fn_meta_group::AbstractString = "group", file_format_expr::AbstractString = "read_mtx", T::Type = Int32, feature_col::Int = 2, barcode_col::Int = 1, feature_threshold::Int = 30, cell_threshold::Int = 200, fn_feature_gene_sit::Int = 1, fn_feature_delim::AbstractChar = '\t', mode_combine_genepair::AbstractString = "all_gene_combine_pair", mode_use_feature::AbstractString = "custom", remove_zeros::Bool = true, use_testdata::AbstractString = "no", work_dir::AbstractString = "./")
The parameters are:
fn_expr::AbstractString: Expression matrix file path, with rows representing genes and columns representing samples, does not require column names. (Default: "matrix.mtx".rn_expr::AbstractString: Gene list file path. (Default: "features.tsv".cn_expr::AbstractString: Sample name list file path. (Default: "barcodes.tsv".fn_meta::AbstractString: Metadata file path, the first column sample name, the second column group information. (Default = "fn_meta.txt".fn_feature::AbstractString: List of characteristic genes (optional). Default: = "fn_feature.txt".building_random_forests::AbstractString: Whether to establish a random forest model. Default = "yes".n_top::Int: The top n_top genes with the highest scores. Default: 1.n_train::Int: Number of training set samples. Default: 13.n_test::Int: Number of test set samples. Default: 3.t_hill_iter_num::Int: Hill climbing number of iterations. Default: 500.t_train_iter_num::Int: The number of iterations of the climb training. Default: 15.ncell_pseudo: pseudo-bulk mode, which combinesncell_pseudocells into one sample. If 0 is set, this mode is disabled Default: 0.fn_meta_delim::AbstractChar: Delimiter of the metadata file. Default: = '\t'.fn_meta_group::AbstractString: Specifies the column name of the group information in metadata. Default: "group".file_format_expr: There are two input modes "read_mtx" and "read_expr_matrix" for the expression profile file format. Default: "read_mtx".T::Type: Datatype in the MTX file. Default: Int32.feature_col::Int: which column is used as feature names. Default: 1 (first).barcode_col::Int: which column is used as barcode names. Default: 1 (first).feature_threshold::Int: the least number of cells that a feature must express in, in order to be kept. Default: 30.cell_threshold::Int: the least number of genes that a cell must express, in order to be kept. Default: 200.fn_feature_gene_sit::Int: The column in the feature gene file where the feature gene resides. Default: 1.fn_feature_delim::AbstractChar: The separator of the characteristic gene file. Default: '\t'.mode_combine_genepair::AbstractString: Gene matching method: "all_gene_combine_pair", which means all gene pairs, "all_feature_gene_combine_pair" indicates the pairing of the feature gene with all genes except itself. "feature_gene_combine_pair" indicates the pairing of feature genes. Default: "all_gene_combine_pair".mode_use_feature::AbstractString: For the use of feature genes, "no" indicates that no feature genes are used, "custom" indicates that custom feature genes are used, and "DEGs_by_RankCompV3" indicates that DEGs (FDR < 0.05) as a characteristic gene. Default: "custom".remove_zeros::Bool: # Filter values that are all zeros. Default: true.use_testdata::AbstractString: Whether to use test data. "yes "or "no". Default: "no".work_dir::AbstractString: Working directory. Default: "./".
Read in the common 10X single-cell RNA expression file in the MTX format (unzipped).
read_mtx(fn::AbstractString, rn::AbstractString, cn::AbstractString; T::Type = Int32, feature_col::Int = 2, barcode_col::Int = 1)
The parameters are:
fn::AbstractString: MTX file path.rn::AbstractString: features file path.cn::AbstractString: barcodes file path.T::Type: Datatype in the MTX file. Default: Int32.feature_col::Int: which column is used as feature names. Default: 1 (first).barcode_col::Int: which column is used as barcode names. Default: 1 (first).
Read in an expression matrix stored in fn where its row names are stored in rn and column names are stored in cn.
It returns (matrix, vector of row names, vector of column names)
read_expr_matrix(fn::AbstractString,rn::AbstractString, cn::AbstractString)
The parameters are:
fn::AbstractString: Expression matrix file path, with rows representing genes and columns representing samples, does not require column names.rn::AbstractString: Gene list file path.cn::AbstractString: Sample name list file path.
Read in a meta data file with the first row assumed to be the header and the row names assumed to be the profile names (cell barcodes).
Grouping information is specified by the column with the header name of group. If group is not found, the second column will be used.
It returns the grouped profile names (vector of vectors) and group names.
read_meta(fn::AbstractString, group::AbstractString = "group"; delim::AbstractChar = '\t')
The parameters are:
fn::AbstractString: Metadata file path, the first column sample name, the second column group information. (Default = "fn_meta.txt".group::AbstractString: Specifies the column name of the group information in metadata. (Default: "group".delim::AbstractChar: Delimiter of the metadata file. (Default: = '\t')
Filter an expression matrix mat, only keep those genes expressed in greater than feature_threshold cells and cells expressing greater than cell_threshold features.
Return the filtered matrix and the bit vectors for keeping features and cells.
filter_expr_matrix(mat::AbstractMatrix, feature_threshold::Int=30, cell_threshold::Int=200)
The parameters are:
mat::AbstractMatrix: expression matrix (either dense or sparse).feature_threshold::Int: the least number of cells that a feature must express in, in order to be kept. Default: 30.cell_threshold::Int: the least number of genes that a cell must express, in order to be kept. Default: 200.
Generate a matrix of pseudobulk profiles from mat which stores single-cell RNA profiles. Each column represents a cell's profile. Each pseudobulk profile is generated from np (default: 10) single-cell profiles.
generate_pseudobulk_kernel(mat::AbstractMatrix, np::Int = 10)
The parameters are:
mat::AbstractMatrix: Each column is a profile.np::Int: Number of profiles in each pseudobulk profile. Default: 10.
Molecular marker gene pairs are obtained by this function:
run_REO_molecular_signatures(nmat::AbstractMatrix, fea::Vector, ngrp::BitVector, bar::Vector; fn_stem::AbstractString, t_hill_iter_num::Int = 500, n_train::Int = 13, n_test::Int = 3, t_train_iter_num::Int = 15, n_top::Int = 10, fn_feature::AbstractString = "fn_feature.txt", mode_combine_genepair::AbstractString = "all_gene_combine_pair", mode_use_feature::AbstractString = "custom", fn_feature_gene_sit::Int = 1, fn_feature_delim::AbstractChar = '\t')
The parameters are:
nmat::AbstractMatrix: Expression profiles matrix.fea::Vector: Gene list.ngrp::BitVector: Sample grouping information, 1 or 0.bar::Vector: Sample group name.fn_stem::AbstractString: Filename.t_hill_iter_num::Int: Hill climbing number of iterations. Default: 500.n_train::Int: Number of training set samples. Default: 13.n_test::Int: Number of test set samples. Default: 3.t_train_iter_num::Int: The number of iterations of the climb training. Default: 15.n_top::Int: The top n_top genes with the highest scores. Default: 10.fn_feature::AbstractString: Feature file name (optional). Default: "fn_feature.txt".mode_combine_genepair::AbstractString: "all_gene_combine_pair", "all_feature_gene_combine_pair", "feature_gene_combine_pair". Default: "all_gene_combine_pair".mode_use_feature::AbstractString: "no", "custom", "DEGs_by_RankCompV3". Default: "custom".fn_feature_gene_sit::Int: The column in which the feature gene is located. Default: 1.fn_feature_delim::AbstractChar: The separator of the characteristic gene file. Default: '\t'.
Train a random forest model.
run_REO_RandomForest(mat::AbstractMatrix, fea::Vector, Y::DataFrame, marker_genepair::Matrix, fn_stem::String)
The parameters are:
nmat::AbstractMatrix: Expression profiles matrix.fea::Vector: Gene list.fea::Vector: Sample grouping information, 1 or 0.marker_genepair::Matrix: The list of marker gene pairs consists of two columns, the first is gene a, the second is gene b, and the relationship between them is a > b.fn_stem::String: File name prefix.