Package 'bmass'

Bayesian multivariate analysis of summary statistics (bmass)

Description

Run bmass on a set of phenotypes that each have univariate GWAS statistics on the same set of SNPs

Usage

bmass(DataSources, GWASsnps = NULL,
  SNPMarginalUnivariateThreshold = 1e-06,
  SNPMarginalMultivariateThreshold = 1e-06, GWASThreshFlag = TRUE,
  GWASThreshValue = 5e-08, NminThreshold = 0,
  PrintMergedData = FALSE, PrintProgress = FALSE, ...)

Arguments

DataSources	A string indicating the variable names of the input datafiles and phenotypes. No default value.
GWASsnps	A data.table containing rows of SNPs that were univariate genome-wide significant in the phenotypes being used for analysis; GWASsnps input file should have two columns, one for chromosome and another for basepair position (with column headers of Chr and BP). No default value.
SNPMarginalUnivariateThreshold	A numerical value indicating the univariate p-value threshold to use when collecting marginally significant SNPs for final bmass analysis. Default is 1e-6.
SNPMarginalMultivariateThreshold	A numerical value indicating the basic multivariate p-value threshold to use when collecting marginally significant SNPs for final bmass analysis. Default is 1e-6.
GWASThreshFlag	A logical TRUE/FALSE flag that indicates whether to threshold input GWASsnps list by a univariate GWAS p-value or not (eg the input GWASsnps list contains variants that are significant from discovery + replication data, but the input summary statistics are just from the discovery cohort). Default is TRUE.
GWASThreshValue	A numerical value indicating the univariate p-value threshold to use in conjunction with the GWASThreshFlag. Default is 5e-8.
NminThreshold	A numerical value that indicates a sample size threshold to use where SNPs below which are removed. Default is 0.
PrintMergedData	A logical TRUE/FALSE flag that indicates whether the intermediary 'merged datafile' should be included in the final bmass output; this file combines all the phenotypes for every SNP provided just prior to thresholding for marginally significant SNPs. Default is FALSE.
PrintProgress	A logical TRUE/FALSE flag that indicates whether progress statements should be printed to stderr during the course of running bmass or not. Default is FALSE.
...	Additional optional arguments.

Value

A list containing model, SNP, and posterior information for both the previously significant univariate SNPs (PreviousSNPs) and the newly significant multivariate SNPs (NewSNPs). For a full breakdown of the bmass output list structure, please see the associated vignettes.

Other Examples

bmass(c("HDL","LDL","TG","TC"), GWASsnps, NminThreshold = 50000) bmass(c("HDL","LDL","TG","TC"), GWASsnps, GWASThreshValue = 1e-8, NminThreshold = 50000, PrintProgress = TRUE) bmass(c("HDL", "LDL", "TG", "TC"), GWASsnps, GWASThreshFlag = FALSE, SNPMarginalUnivariateThreshold = 1e-4, SNPMarginalMultivariateThreshold = 1e-4, PrintMergedData = TRUE) bmassOutput <- bmass(c("HDL","LDL","TG","TC"), GWASsnps, NminThreshold = 50000)

Examples

Phenotypes <- c("bmass_SimulatedData1", "bmass_SimulatedData2")
bmassOutput <- bmass(Phenotypes, bmass_SimulatedSigSNPs)
summary(bmassOutput)
bmassOutput$NewSNPs$SNPs

---

bmass Simulated Dataset 1

Description

A manually created sample dataset for use in Roxygen2 documents and vignettes.

Format

A data frame with 11 rows and 9 variables:

Chr

chromosome

BP

basepair position

Marker

rsID# or other identifier

MAF

Minor Allele Frequency

A1

reference allele

A2

alternative allele

Direction

direction of association effect size, + or -

pValue

p-Value of GWAS association

N

sample size

Source

Manually created

---

bmass Simulated Dataset 2

Description

A manually created sample dataset for use in Roxygen2 documents and vignettes.

Format

A data frame with 11 rows and 9 variables:

Chr

chromosome

BP

basepair position

Marker

rsID# or other identifier

MAF

Minor Allele Frequency

A1

reference allele

A2

alternative allele

Direction

direction of association effect size, + or -

pValue

p-Value of GWAS association

N

sample size

Source

Manually created

---

bmass Simulated GWAS SNPs

Description

A manually created list of GWAS significant SNPs to be used in conjunction with 'bmass_SimulatedData1' and 'bmass_SimulatedData2'.

Format

A data frame with 2 rows and 2 variables:

Chr

chromosome

BP

basepair position

Source

Manually created

---

bmass Test Dataset 1

Description

A manually created sample dataset for use in unit tests.

Format

A data frame with 11 rows and 9 variables:

Chr

chromosome

BP

basepair position

Marker

rsID# or other identifier

MAF

Minor Allele Frequency

A1

reference allele

A2

alternative allele

Direction

direction of association effect size, + or -

pValue

p-Value of GWAS association

N

sample size

Source

Manually created

---

bmass Test Dataset 2

Description

A manually created sample dataset for use in unit tests.

Format

A data frame with 11 rows and 9 variables:

Chr

chromosome

BP

basepair position

Marker

rsID# or other identifier

MAF

Minor Allele Frequency

A1

reference allele

A2

alternative allele

Direction

direction of association effect size, + or -

pValue

p-Value of GWAS association

N

sample size

Source

Manually created

---

bmass Test GWAS SNPs

Description

A manually created list of GWAS significant SNPs to be used in conjunction with 'bmass_TestData1' and 'bmass_TestData2'.

Format

A data frame with 2 rows and 2 variables:

Chr

chromosome

BP

basepair position

Source

Manually created

---

Get Marginal {U,D,I} Posteriors

Description

Get marginal posteriors for how much every individual phenotype belongs to categories {U,D,I} across each SNP

Usage

GetMarginalPosteriors(DataSources, ListSNPs, Models, LogFile)

Arguments

DataSources	A string indicating the variable names of the input datafiles and phenotypes.
ListSNPs	A list produced from running bmass containing the SNPs of interest to get marginal posteriors for.
Models	A matrix describing the models being explored (default output from running bmass).
LogFile	A matrix of string outputs for function logging purposes (default output from running bmass).

Value

A list containing three matrices of SNPs x Phenotypes marginal posteriors for each category {U,D,I}; this list is appended to the input ListSNPs as a new object, Marginals (the full returned object is a list containing the input ListSNPs and the input LogFile).

Examples

Phenotypes <- c("bmass_SimulatedData1", "bmass_SimulatedData2")
bmassOutput <- bmass(Phenotypes, bmass_SimulatedSigSNPs)
bmassOutput[c("PreviousSNPs", "LogFile")] <-
GetMarginalPosteriors(Phenotypes, bmassOutput$PreviousSNPs,
bmassOutput$Models, bmassOutput$LogFile)
bmassOutput$PreviousSNPs$Marginals

---

Get Model Prior Matrix

Description

Creates a matrix containing the model descriptions and their associated priors.

Usage

GetModelPriorMatrix(DataSources, Models, ModelPriors, LogFile,
  SigmaAlphas = c(0.005, 0.0075, 0.01, 0.015, 0.02, 0.03, 0.04, 0.05,
  0.06, 0.07, 0.08, 0.09, 0.1, 0.15))

Arguments

DataSources	A string indicating the variable names of the input datafiles and phenotypes.
Models	A matrix describing the models being explored (default output from running bmass).
ModelPriors	A vector containing the priors on each model across each tranche of sigma alpha (default output from running bmass; length is number of models times number of sigma alphas).
LogFile	A matrix of string outputs for function logging purposes (default output from running bmass).
SigmaAlphas	A vector containing the different values traversed for this 'effect size controlling' hyperparameter (see "Prior on Sigma_Alpha" in Stephens 2013 PLoS ONE, https://doi.org/10.1371/journal.pone.0065245).

Value

A matrix containing the original description of each model sort by prior, each model's trained prior, the cummulative prior distribution, and the model's original order position.

Examples

Phenotypes <- c("bmass_SimulatedData1", "bmass_SimulatedData2")
bmassOutput <- bmass(Phenotypes,bmass_SimulatedSigSNPs)
bmassOutput[c("ModelPriorMatrix", "LogFile")] <- 
  GetModelPriorMatrix(Phenotypes, bmassOutput$Models,
  bmassOutput$ModelPriors, bmassOutput$LogFile)
head(bmassOutput$ModelPriorMatrix)

---

Get Top Multivariate Models

Description

Get a summary of the top models per SNP across all multivariate {U,D,I} combinations based on posterior probabilities.

Usage

GetTopModelsPerSNPViaPosteriors(DataSources, ListSNPs, ModelPriorMatrix,
  LogFile)

Arguments

DataSources	A string indicating the variable names of the input datafiles and phenotypes.
ListSNPs	A list produced from running bmass containing the SNPs of interest to get marginal posteriors for.
ModelPriorMatrix	A matrix detailing the models being explored and their associated priors (obtained by running GetModelPriorMatrix)
LogFile	A matrix of string outputs for function logging purposes (default output from running bmass).

Value

A matrix containing each model that was a SNP's top model at least once, along with related information; this matrix is appended to the input ListSNPs as a new object, TopModels (the full returned object is a list containing the input ListSNPs and the input LogFile).

Examples

Phenotypes <- c("bmass_SimulatedData1", "bmass_SimulatedData2")
bmassOutput <- bmass(Phenotypes, bmass_SimulatedSigSNPs)
bmassOutput[c("ModelPriorMatrix", "LogFile")] <- 
  GetModelPriorMatrix(Phenotypes, bmassOutput$Models,
  bmassOutput$ModelPriors, bmassOutput$LogFile)
bmassOutput[c("PreviousSNPs", "LogFile")] <-
  GetTopModelsPerSNPViaPosteriors(Phenotypes,
  bmassOutput$PreviousSNPs, bmassOutput$ModelPriorMatrix, bmassOutput$LogFile)
head(bmassOutput$PreviousSNPs$TopModels)

---

GlobalLipids2013 GWAS SNPs

Description

A list of the univariate GWAS significant SNPs from the GlobalLipids2013 dataset to be used in the second introductory bmass vignette.

Format

A data frame with 157 rows and 2 variables:

Chr

chromosome

BP

basepair position

Source

Supplementary Tables 2 and 3 from https://doi.org/10.1038/ng.2797.

---