BOOST.GP: Fit BOOST-GP Model for Gene Expression Levels

In estfernan/boost: Bayesian Modeling of Spatial Transcriptomics Data

Fit BOOST-GP Model for Gene Expression Levels

Description

Fit the BOOST-GP model to detect whether the gene is spatially variable (SV). The fit is done within a Metropolis (SSVS) search variable selection algorithm. Only one gene must be present, but no normalization is necessary.

Usage

BOOST.GP(
  abs.expr,
  spots,
  size.factor = NULL,
  gene.name = NULL,
  n.iter = 1000,
  burn.prop = 0.5,
  update.prop = 0.2,
  init.b.sigma = NULL,
  init.h = 1
)

Arguments

abs.expr	A numeric vector p of length n that denotes the absolute gene expression levels. Each entry is an integer that denotes the gene count at spot i.
spots	An n-by-2 numeric matrix T to represent the geospatial profile, where each row indicates the spot location in the grid.
size.factor	A numeric vector s of length n to compute the relative gene expression levels. Each entry denotes the size factor of sample i that captures all nuisance effects. The default is NULL to evaluate the absolute expression levels.
gene.name	A character string that specifies the name of the gene passed. To be used when storing the results. The default value is NULL to keep the gene expression levels unnamed.
n.iter	An integer value to specify the number of iterations for the DMH algorithm. The default is 1,000 iterations.
burn.prop	A numeric value to specify the proportion of iterations to use as warm-up. The default is 0.50 to use half of the iterations for warm-up.
update.prop	A numeric value to specify the proportion of samples to update in each iteration. The default is 0.2 to update one-fifth of the total samples.
init.b.sigma	A numeric value to specify the initial value of the scale parameter in the inverse-gamma prior on the variance of the multivariate normal distribution prior for the log-expression levels. The default is NULL to set the value as twice the sample variance of the log-expression levels.
init.h	A numeric value to specify the scaling of the variance for the normal prior set on each coefficient. The default is one to not scale the variance.

Details

The primary interest lies in the identification of SV genes via a selection indicator. See Li et al. (2020) for more information on the model fitting and posterior inference procedures.

Value

BOOST.GP returns an object of class "BOOST.GP". The function base::print() i.e., print.BOOST.GP(), can be used to print a summary of the results.

An object of class "BOOST.GP" is a list containing the following components:

call	the function call in which all of the specified arguments are specified by their full names.
model	the name of statistical model or technique.
gene.name	the name of gene evaluated.
summary	a summary table that contains a summary of the estimated parameters.
measures	the estimated Bayes factor and corresponding p-value
time	the execution time of the function.

References

Li, Q., Zhang M., Xie Y., & Xiao, G. (2020). Bayesian Modeling of Spatial Molecular Profiling Data via Gaussian Process. arXiv preprint arXiv:2012.03326.

See Also

get.size.factor() for estimating the size factor; print.BOOST.GP() for printing a summary of results to console.

Examples

## Not run: 
library(boost)

## load sample dataset
data(mob)

## extract a sample gene and get size factor
g <- mob[, "Apoe"]
s <- get.size.factor(mob, estimation.method = "TSS")

## fit the model
res <- BOOST.GP(g, mob.spots, size.factor = s, gene.name = "Apoe")
print(res)

## End(Not run)

estfernan/boost documentation built on June 24, 2022, 12:20 a.m.