PCLasso: Protein complex-based group lasso-Cox model

In weiliu123/PCLasso: A protein complex-based group lasso-Cox model for accurate prognosis and risk protein complex discovery

Description

Construct a PCLasso model based on a gene expression matrix, survival data, and protein complexes.

Usage

PCLasso(x, y, group, penalty = c("grLasso", "grMCP", "grSCAD", "gel", "cMCP"), 
standardize = TRUE, ...)

Arguments

x	A n x p matrix of gene expression measurements with n samples and p genes.
y	The time-to-event outcome, as a two-column matrix or Surv object. The first column should be time on study (follow up time); the second column should be a binary variable with 1 indicating that the event has occurred and 0 indicating (right) censoring.
group	A list of groups. The feature (gene) names in group should be consistent with the feature (gene) names in x.
penalty	The penalty to be applied to the model. For group selection, one of grLasso, grMCP, or grSCAD. For bi-level selection, one of gel or cMCP. See grpsurv in the R package grpreg for details.
standardize	Logical flag for x standardization, prior to fitting the model. Default is TRUE.
...	Arguments to be passed to grpsurv in the R package grpreg.

Details

The PCLasso model is a prognostic model which selects important predictors at the protein complex level to achieve accurate prognosis and identify risk protein complexes. The PCLasso model has three inputs: a gene expression matrix, survival data, and protein complexes. It estimates the correlation between gene expression in protein complexes and survival data at the level of protein complexes. Similar to the traditional Lasso-Cox model, PCLasso is based on the Cox PH model and estimates the Cox regression coefficients by maximizing partial likelihood with regularization penalty. The difference is that PCLasso selects features at the level of protein complexes rather than individual genes. Considering that genes usually function by forming protein complexes, PCLasso regards genes belonging to the same protein complex as a group, and constructs a l1/l2 penalty based on the sum (i.e., l1 norm) of the l2 norms of the regression coefficients of the group members to perform the selection of features at the group level. Since a gene may belong to multiple protein complexes, that is, there is overlap between protein complexes, the classical group Lasso-Cox model for non-overlapping groups may lead to false sparse solutions. The PCLasso model deals with the overlapping problem of protein complexes by constructing a latent group Lasso-Cox model. And by reconstructing the gene expression matrix of the protein complexes, the latent group Lasso-Cox model is transformed into a non-overlapping group Lasso-Cox model in an expanded space, which can be directly solved using the classical group Lasso method. Through the final sparse solution, we can predict the patient's risk score based on a small set of protein complexes and identify risk protein complexes that are frequently selected to construct prognostic models.

Value

An object with S3 class "PCLasso" containing:

fit	An object of class "grpsurv"
group.dt	Groups with features (genes) not included in x being filtered out.

Author(s)

Wei Liu

References

PCLasso: a protein complex-based group lasso-Cox model for accurate prognosis and risk protein complex discovery. To be published.

Park, H., Niida, A., Miyano, S. and Imoto, S. (2015) Sparse overlapping group lasso for integrative multi-omics analysis. Journal of computational biology: a journal of computational molecular cell biology, 22, 73-84.

See Also

predict.PCLasso, cv.PCLasso

Examples

library("survival")

# load data
data(ExpMatrix)
data(survData)
data(PCGroup)

x <- ExpMatrix    
y <- Surv(time=survData[,"time"], event=survData[,"status"])

# fit the PCLasso model
fit1 <- PCLasso(x, y, group = PCGroup, penalty = "grLasso")

weiliu123/PCLasso documentation built on March 28, 2021, 3:51 a.m.