calculate_SP: Calculate starting points to be used in the likelihood...

In LizaDarrous/lhcMR: Latent Heritable Confounder - Mendelian Randomisation

Calculate starting points to be used in the likelihood function optimisation

Description

Calculate starting points to be used in the likelihood function optimisation

Usage

calculate_SP(
  input.df,
  trait.names,
  run_ldsc = TRUE,
  run_MR = TRUE,
  saveRFiles = TRUE,
  hm3 = NA,
  ld = NA,
  nStep = 2,
  SP_single = 3,
  SP_pair = 50,
  SNP_filter = 10,
  SNP_filter_ldsc = NA,
  nCores = 1,
  M = 1e+07
)

Arguments

input.df	The resulting data frame from merge_sumstats(), where the effect size, SE, RSID and other columns are present, in addition to columns representing LD scores, weights and local LD structure
trait.names	Vector containing the trait names in the order they were used in merge_sumstats(): Exposure, Outcome
run_ldsc	Boolean. Whether GenomicSEM::ldsc should be run to obtain the cross trait-intercept (i_XY). If FALSE, a random value will be generated. Default value = TRUE
run_MR	Boolean. Whether TwoSampleMR::mr should be run to obtain the bidirectional causal effects (axy_MR, ayx_MR). If FALSE, random values will be generated. Default value = TRUE
saveRFiles	Boolean, whether to write the results of GenomicSEM::ldsc,TwoSampleMR::mr, and the single trait analysis of LHC-MR (returns trait intercept and polygenicity) Default value = TRUE
hm3	Path to the input file (HAPMAP3 SNPs) required by GenomicSEM::ldsc
ld	Path to the input file (LD scores) required by GenomicSEM::ldsc
nStep	Can take two numerical values: 1 or 2. Represents the number of steps the lhcMR analysis will undertake. One single step estimates all 9 parameters simultaneously while fixing only the traits' intercepts iX and iY, while two steps estimates 7 parameters after having estimated traits' intercepts and polygenicity (iX, piX, iY, piY) from the single trait analysis and fixed their values in the likelihood optimisation and parameter estimation
SP_single	Numerical value indicating how many starting points should the single trait analysis use in the likelihood optimisation. Best to range between 3-5, default value = 3
SP_pair	Numerical value indicating how many starting points should the pair trait analysis use in the likelihood optimisation. Best to range between 50-100, default value = 50
SNP_filter	Numerical value indicating the filtering of every nth SNP to reduce large datasets and speed up analysis. Default value = 10
SNP_filter_ldsc	Numerical value indicating the filtering of every nth SNP to reduce large datasets and speed up the LDSC analysis. Set to 1 if no filtering is needed, otherwise default = 10
nCores	Numerical value indicating number of cores to be used in 'mclapply' to parallelise the analysis. If set to NA, then it will be calculated as 2/3 of the available cores, default value = 1 to avoid parallelisation
M	Numerical value indicating the number of SNPs used to calculate the LD reported in the LD file (for genotyped SNPs). Default value = 1e7

Value

Returns a list containing the filtered dataset (by every SNP_filterth SNP), the starting points to be used in the pair trait optimisation, the traits' intercepts, the traits' polygenicity if nStep = 2, as well as some extra parameters like the cross-trait intercept and bidirectional causal effect estimated by IVW

LizaDarrous/lhcMR documentation built on March 24, 2024, 10:28 p.m.