getspres {getspres}

Exploring Heterogeneity in Meta-Analysis with SPRE Statistics.

Description

getspres computes SPRE (standardised predicted random-effects) statistics to identify outlier studies in genetic association meta-analyses which might have undue influence on the average genetic effect leading to inflated genetic signals.

Usage

getspres(beta_in, se_in, study_names_in, variant_names_in, ...)

## Default S3 method:
getspres(
  beta_in,
  se_in,
  study_names_in,
  variant_names_in,
  tau2_method = "DL",
  verbose_output = FALSE,
  ...
)

Arguments

beta_in

A numeric vector of study effect-sizes e.g. log odds-ratios.

se_in

A numeric vector of standard errors, genomically corrected at study-level.

study_names_in

A character vector of study names.

variant_names_in

A character vector of variant names e.g. rsIDs.

...

other arguments.

tau2_method

A character scalar, specifying the method that should be used to estimate heterogeneity either through DerSimonian and Laird's moment-based estimate "DL" or restricted maximum likelihood "REML". Note: The REML method uses the iterative Fisher scoring algorithm (step length = 0.5, maximum iterations = 10000) to estimate tau2. Default is "DL".

verbose_output

An optional boolean to display intermediate output. (Default is FALSE).

Details

SPRE statistics are precision-weighted residuals that summarise the direction and extent with which observed study effects in a meta-analysis differ from the summary (or average genetic) effect. See the getspres website for more information, documentation and examples.

getspres takes as input study effect-size estimates and their corresponding standard errors (i.e. summary data). Study effect estimates could be in the form of linear regression coefficients or log-transformed regression coefficients (per-allele log odds ratios) from logistic regression.

getspres uses inverse-variance weighted meta-analysis models in the metafor R package to calculate SPRE statistics.

Value

Returns a list containing:

• number_variants A numeric scalar for the number of variants

• number_studies A numeric scalar for the number of studies

• spre_dataset A dataframe that is a dataset of computed SPRE statistics and contains the following fields:

  • beta , study effect-size estimates

  • se , corresponding standard errors of the study effect-size estimates

  • variant_names , variant names

  • study_names , study names

  • study , study numbers

  • snp , snp numbers

  • tau2 , tau_squared, estimate of amount of between-study variance

  • I2 , I_squared, heterogeneity index (Higgins inconsistency metric) representing proportion of total observed variation due to between-study variance

  • Q , Q-statistic (Cochran's Q)

  • xb , prediction excluding random effects

  • xbse , standard error of prediction excluding random effects

  • xbu , predictions including random effects

  • stdxbu , standard error of prediction (fitted values) including random effects

  • hat , leverage a.k.a diagonal elements of the projection hat matrix

  • rawresid , raw residuals

  • uncondse , unconditional standard errors

  • spre , SPRE statistics (standardised predicted random effects) i.e. raw residuals divided by the unconditional standard errors

Methods (by class)

• default: Computes SPRE statistics in genetic association meta-analyses

See Also

https://magosil86.github.io/getspres/ to the visit getspres website.

Examples

library(getspres)


# Calculate SPRE statistics for a subset of variants in the heartgenes214 dataset.
# heartgenes214 is a case-control GWAS meta-analysis of coronary artery disease.
# To learn more about the heartgenes214 dataset ?heartgenes214

# Calculating SPRE statistics for 3 variants in heartgenes214

heartgenes3 <- subset(heartgenes214, 
    variants %in% c("rs10139550", "rs10168194", "rs11191416")) 

getspres_results <- getspres(beta_in = heartgenes3$beta_flipped, 
                               se_in = heartgenes3$gcse, 
                      study_names_in = heartgenes3$studies, 
                    variant_names_in = heartgenes3$variants)


# Explore results generated by the getspres function
str(getspres_results)

# Retrieve number of studies and variants
getspres_results$number_variants
getspres_results$number_studies

# Retrieve SPRE dataset
df_spres <- getspres_results$spre_dataset
head(df_spres)

# Extract SPREs from SPRE dataset
head(spres <- df_spres[, "spre"])



# Exploring available options in the getspres function:
#     1. Estimate heterogeneity using "REML", default is "DL"
#     2. Calculate SPRE statistics verbosely
getspres_results <- getspres(beta_in = heartgenes3$beta_flipped, 
                               se_in = heartgenes3$gcse, 
                      study_names_in = heartgenes3$studies, 
                    variant_names_in = heartgenes3$variants,
                         tau2_method = "REML",
                      verbose_output = TRUE)

str(getspres_results)