runDEAnalysis: Perform differential expression analysis and identification...

In YosefLab/ImpulseDE2: Differential expression analysis of longitudinal count data sets

Perform differential expression analysis and identification of transiently activated or deactivated genes.

Description

Performs model selction based on loglikelihood ratio tests. The primary model selection is the differential expression analysis. The secondary model selection is the selection between a sigmoidal and an impulse fit for differentially expressed genes which is used to define transiently activated or deactivated genes.

Usage

runDEAnalysis(objectImpulseDE2, boolCaseCtrl, boolIdentifyTransients,
  scaQThresTransients = 0.001)

Arguments

objectImpulseDE2	(object class ImpulseDE2Object) Object containing fits to be evaluated.
boolCaseCtrl	(bool) Whether to perform case-control analysis. Does case-only analysis if FALSE.
boolIdentifyTransients	(bool) [Defaul FALSE] Whether to identify transiently activated or deactivated genes. This involves an additional fitting of sigmoidal models and hypothesis testing between constant, sigmoidal and impulse model.
scaQThresTransients	(scalar) [Default 0.001] FDR-corrected p-value threshold for hypothesis tests between impulse, sigmoidal and constant model used to identify transiently regulated genes.

Value

(ImpulseDE2Object) Input object with dfDEAnalysis updated to: dfDEAnalysis (data frame samples x reported characteristics) Summary of fitting procedure and differential expression results for each gene.

• Gene: Gene ID.

• p: P-value for differential expression.

• padj: Benjamini-Hochberg false-discovery rate corrected p-value for differential expression analysis.

• loglik_full: Loglikelihood of full model.

• loglik_red: Loglikelihood of reduced model.

• df_full: Degrees of freedom of full model.

• df_red: Degrees of freedom of reduced model

• mean: Inferred mean parameter of constant model of first batch. From combined samples in case-ctrl.

• allZero (bool) Whether there were no observed non-zero observations of this gene. If TRUE, fitting and DE analsysis were skipped and entry is NA.

Entries only present in case-only DE analysis:

• converge_impulse: Convergence status of optim for impulse model fit (full model).

• converge_const: Convergence status of optim for constant model fit (reduced model).

Entries only present in case-control DE analysis:

• converge_combined: Convergence status of optim for impulse model fit to case and control samples combined (reduced model).

• converge_case: Convergence status of optim for impulse model fit to samples of case condition (full model 1/2).

• converge_control: Convergence status of optim for impulse model fit to samples of control condition (full model 2/2).

Entries only present if boolIdentifyTransients is TRUE:

• converge_sigmoid: Convergence status of optim for sigmoid model fit to samples of case condition.

• impulseTOsigmoid_p: P-value of loglikelihood ratio test impulse model fit versus sigmoidal model on samples of case condition.

• impulseTOsigmoid_padj: Benjamini-Hochberg false-discovery rate corrected p-value of loglikelihood ratio test impulse model fit versus sigmoid model on samples of case condition.

• sigmoidTOconst_p: P-value of loglikelihood ratio test sigmoidal model fit versus constant model on samples of case condition.

• sigmoidTOconst_padj: Benjamini-Hochberg false-discovery rate corrected p-value of loglikelihood ratio test sigmoidal model fit versus constant model on samples of case condition.

• isTransient (bool) Whether gene is transiently activated or deactivated and differentially expressed.

• isMonotonous (bool) Whether gene is not transiently activated or deactivated and differentially expressed. This scenario corresponds to a montonous expression level increase or decrease.

Author(s)

David Sebastian Fischer

See Also

Called by runImpulseDE2.

YosefLab/ImpulseDE2 documentation built on Sept. 17, 2022, 2:45 a.m.