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", "LP_SGL", "PIPET",
    "rSIDISH", "SCIPAC"),
  ...
)

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 matrix/Matrix (genes x cells) or a Seurat object containing scRNA-seq data to be screened.

phenotype

Phenotype data, either: - Named vector (names match matched_bulk columns), or - Patient survival Data frame with row names matching matched_bulk columns, colnames named "time" and "status"

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 infomation

screen_method

Screening algorithm to use, there are seven options:

"Scissor": see also DoScissor()
"scPP": see also DoscPP()
"scPAS": see also DoscPAS()
"scAB": see also DoscAB(), continuous phenotype is not supported
"DEGAS": see also DoDEGAS()
"LP_SGL": see also DoLP_SGL()
"PIPET": see also DoPIPET()
"SIDISH": see also DoSIDISH()

...

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 0.005
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"

SIDISH

sidish_params: (list) SIDISH parameters, default "list()"
env_params: (list) Environment parameters for SIDISH, default "list()"

LP_SGL

resolution

(numeric) Resolution parameter for Leiden clustering, default 0.6

alpha

(numeric) Alpha parameter for SGL balancing L1 and L2 penalties, default 0.5

nfold

(integer) Number of folds for cross-validation, default 5

dge_analysis

(list) Differential expression analysis settings:

run: (logical) Whether to run DEG analysis, default FALSE
logFC_threshold: (numeric) Log fold change threshold, default 1
pval_threshold: (numeric) P-value threshold, default 0.05

PIPET

group

(character or NULL) Name of a metadata column (e.g., "orig.ident") to stratify cells before screening. When NULL (default), screening is performed globally across all cells.

discretize_method

(character) Strategy to binarize continuous phenotypes internally before marker identification. One of:

"median" (default): Equivalent to 2-quantile split (i.e., median threshold).
"kmeans": Two-cluster k-means on the continuous phenotype.
"custom": User-defined cutoffs via cutoff.

cutoff

(numeric vector or NULL) Required only if discretize_method = "custom". Specifies interior breakpoints on the normalized, log2-transformed phenotype scale (i.e., after scale(log2(x + 1))). Must be sorted ascending and of length n_group - 1.

label_type

(character) Phenotype label type (e.g., "PIPET_SBS1"), stored in scRNA_data@misc. Default: "PIPET".

log2FC

(numeric) Absolute log2 fold-change cutoff for differential expression marker selection in bulk data (via DESeq2-like analysis). Default: 1.

p_adjust

(numeric) Adjusted p-value (FDR) cutoff for marker gene selection. Default: 0.05.

show_log2FC

(logical) Whether to annotate markers with signed log2FC direction (e.g., CD3D_up). Default: TRUE.

freq_counts

(integer or NULL) Minimum number of cells a gene must be expressed in to be retained in scRNA-seq data preprocessing. Default: NULL (no filtering).

normalize

(logical) Whether to apply log-normalization (LogNormalize) to scRNA-seq counts prior to correlation. Default: TRUE.

scale

(logical) Whether to scale (center + unit-variance) gene expression across cells before computing distances. Default: TRUE.

nPerm

(integer) Number of label permutations to assess significance of correlation scores. Default: 1000.

distance

(character) Distance or similarity metric for template matching. Supported: "cosine" (default), "pearson", "spearman", "kendall", "euclidean", "maximum".

seed

(integer or NULL) Random seed for reproducibility in marker creation and permutation tests. Default: inherits from getFuncOption("seed").

verbose

(logical) Whether to print progress messages. Default: inherits from getFuncOption("verbose").

parallel

(logical) Whether to enable parallel permutations (requires future::plan() pre-set). Default: FALSE.

Value

A list containing:

scRNA_data: A Seurat object with phenotype-associated cells labelled in meta.data column
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
LP_SGL	All three types	resolution, alpha, nfold, dge_analysis
PIPET	Binary/Continuous	group, discretize_method, cutoff, log2FC, p_adjust, show_log2FC, freq_counts, normalize, scale, nPerm, distance
SIDISH	Survival Only	sidish_params, env_params

Usage

Arguments

Value

Data Matching Requirements

Method Compatibility

See also