MiXcan: The core function of MiXcan package for estimating the...

In JiayiJi/MiXcan:

Description

The core function of MiXcan package for estimating the cell-type specific and non-specific prediction weights of a gene.

Usage

MiXcan(y, x, cov = NULL, pi, nameMatrix = NULL, foldid = NULL)

Arguments

y:	The pre-cleaned expression level data for a single gene in N samples.
x:	A N by P matrix for all the genetic predictors used to predict the expression level of the gene.
cov:	A N by Q matrix for the covariates adjusted in the model (e.g. age, population stratification).
pi:	An estimation of cell-type faction of the cell type of interest (e.g. epithelial). It can be estimated using existing methods in the literature of from the output of pi_estimation function.
nameMatrix:	Default is NULL. A matrix to save the information of variables from the X matrix, such as variable ID, position, rsid, ref_allele, eff_allele.
foldid:	Default is NULL. 10-fold cross-validation (CV) is used in our pipeline. A random split is considered if foldid is NULL. Otherwise foldid is used to split the data for CV.

Value

list with 9 elements. It contains

type:	Whether the prediction model is "CellSpecific" or "NonSpecific.
beta.SNP.cell1:	The prediction weights of the genetic predictors in cell type 1 (the cell type of interest).
beta.SNP.cell2:	The prediction weights of the genetic predictors in cell type 2 (other cell types).
beta.all.models:	All regression coefficients are saved in beta.all.models, including intercepts, coefficients of genetic and non-genetic predictors in cell-type specific and non-specific models.
glmnet.cell:	The cell-type specific prediction model selected using elastic net. This model may not be the final model of MiXcan as elastic net selected parameters may be insignificant.
glmnet.tissue:	The prediction model without considering cell type composition (same as PrediXcan).

JiayiJi/MiXcan documentation built on Dec. 18, 2021, 1:30 a.m.