Single-Cell Data Screening

Integrates matched bulk expression data and phenotype information to identify phenotype-associated cell populations in single-cell RNA-seq data using one of four computational methods. Ensures consistency between bulk and phenotype data before analysis.

Usage

Screen(
  matched_bulk,
  sc_data,
  phenotype,
  label_type = NULL,
  phenotype_class = c("binary", "survival", "continuous"),
  screen_method = c("Scissor", "scPP", "scPAS", "scAB", "DEGAS"),
  ...
)

Arguments

matched_bulk

Matrix or data frame of preprocessed bulk RNA-seq expression data (genes x samples). Column names must match names/IDs in phenotype.

sc_data

A Seurat object containing scRNA-seq data to be screened.

phenotype

Phenotype data, either: - Named vector (names match matched_bulk columns), or - Data frame with row names matching matched_bulk columns

label_type

Character specifying phenotype label type (e.g., "SBS1", "time")

phenotype_class

Type of phenotypic outcome (must be consistent with input data): - "binary": Binary traits (e.g., case/control) - "continuous": Continuous measurements - "survival": Survival objects

screen_method

Screening algorithm to use, there are four options: - "Scissor": see also DoScissor() - "scPP": see also DoscPP() - "scPAS": see also DoscPAS() - "scAB": see also DoscAB(), no continuous support - "DEGAS": see also DoDEGAS()

...

Additional method-specific parameters:

Scissor

alpha: (numeric or NULL) Significance threshold. When NULL, alpha will keep increasing iteratively until the corresponding cells are screened out, default 0.05
cutoff: (numeric) A threshold for terminating the iteration of alpha, only work when alpha is NULL, default 0.2
path2load_scissor_cache: (character) default NULL
path2save_scissor_inputs: (character) A path to save the intermediary data. By using path2load_scissor_cache, the intermediary data can be loaded from the specified path. default "Scissor_inputs.RData"
reliability_test: (logical) Whether to perform reliability test, default FALSE
reliability_test.nfold: (integer) Cross-validation folds for reliability test, default 10
reliability_test.n: (integer) Number of cells to use for reliability test, default 10
cell_evaluation: (logical) Whether to perform cell evaluation, default FALSE
cell_evaluation.benchmark_data: .RData Benchmark data for cell evaluation, default NULL
cell_evaluation.FDR: (numeric) FDR threshold for cell evaluation, default 0.05
cell_evaluation.bootstrap_n: (integer) Number of bootstrap samples for cell evaluation, default 10

scPP

ref_group: (integer or character) Reference group or baseline for binary comparisons, e.g. "Normal" for Tumor/Normal studies and 0 for 0/1 case-control studies. default: 0
Log2FC_cutoff: (numeric) Minimum log2 fold-change for binary markers, default 0.585
estimate_cutoff: (numeric) Effect size threshold for continuous traits, default 0.2
probs: (numeric) Quantile cutoff for cell classification, default 0.2

scPAS

assay: (character) Assay to use from sc_data, default "RNA"
imputation: (logical) Whether to perform imputation, default FALSE
nfeature: (integer) Number of features to select, default 3000
alpha: (numeric or NULL) Significance threshold, When NULL, alpha will keep increasing iteratively until the corresponding cells are screened out, default 0.01
independent: (logical) The background distribution of risk scores is constructed independently of each cell. default: TRUE
network_class: (character) Network class to use. default: 'SC', indicating gene-gene similarity networks derived from single-cell data. The other one is 'bulk'.
permutation_times: (integer) Number of permutations, default 2000
FDR_threshold: (numeric) FDR value threshold for identifying phenotype-associated cells default 0.05

scAB

alpha: (numeric) Coefficient of phenotype regularization ,default 0.005
alpha_2: (numeric) Coefficent of cell-cell similarity regularization, default 5e-05
maxiter: (integer) NMF optimization iterations, default 2000
tred: (integer) Z-score threshold, default 2

DEGAS

sc_data.pheno_colname: (character) Phenotype column name in sc_data, default "NULL"
select_fraction: (numeric) Fraction of cells to select for DEGAS, default 0.05
tmp_dir: (character) Temporary directory for DEGAS, default "NULL"
env_params: (list) Environment parameters for DEGAS, default "list()"
degas_params: (list) DEGAS parameters, default "list()"
normality_test_method: (character) Normality test method for DEGAS, default "jarque-bera"

Value

A list containing:

scRNA_data: Filtered Seurat object with phenotype-associated cells
Some screen_result: Important information about the screened result related to the selected method

Data Matching Requirements

matched_bulk column names and phenotype names/rownames must be identical
Phenotype values must correspond to bulk samples (not directly to single cells)
Mismatches will trigger an error before analysis begins, and there is a built-in pre-run check.

Method Compatibility

Method	Supported Phenotypes	Additional Parameters
`Scissor`	All three types	`alpha`, `cutoff`, `path2load_scissor_cache`, `path2save_scissor_inputs`, `reliability_test`, `reliability_test.n`,`reliability_test.nfold`, `cell_evaluation`,`cell_evaluation.benchmark_data`,`cell_evaluation.FDR`,`cell_evaluation.bootstrap_n`
`scPP`	All three types	`ref_group`, `Log2FC_cutoff`, `estimate_cutoff`, `probs`
`scPAS`	All three types	`n_components` ,`assay`, `imputation`,`nfeature`, `alpha`,`network_class`,`permutation_times`,`FDR_threshold`,`independent`
`scAB`	Binary/Survival	`alpha`, `alpha_2`, `maxiter`, `tred`
`DEGAS`	All three types	`sc_data.pheno_colname`,`select_fraction`,`tmp_dir`,`env_params`,`degas_params`,`normality_test_method`

Usage

Arguments

Value

Data Matching Requirements

Method Compatibility

See also