Skip to contents

Run the SuSiE RSS pipeline

Source: R/susie_wrapper.R
susie_rss_pipeline.Rd

This function runs the SuSiE RSS pipeline, performing analysis based on the specified method. It processes the input summary statistics and LD data to provide results in a structured output.

Usage

susie_rss_pipeline(
  sumstats,
  LD_mat,
  n = NULL,
  var_y = NULL,
  L = 5,
  max_L = 30,
  l_step = 5,
  analysis_method = c("susie_rss", "single_effect", "bayesian_conditional_regression"),
  coverage = 0.95,
  secondary_coverage = c(0.7, 0.5),
  signal_cutoff = 0.1,
  min_abs_corr = 0.8
)

Arguments

sumstats

A list or data frame containing summary statistics with 'z' or 'beta' and 'se' columns.

LD_mat

The LD matrix.

n

Sample size (default: NULL).

var_y

Variance of Y (default: NULL).

L

Initial number of causal configurations to consider in the analysis (default: 5).

max_L

Maximum number of causal configurations to consider in the analysis (default: 30).

l_step

Step size for increasing L when the limit is reached during dynamic adjustment (default: 5).

analysis_method

The analysis method to use. Options are "susie_rss", "single_effect", or "bayesian_conditional_regression" (default: "susie_rss").

coverage

Coverage level for susie_rss analysis (default: 0.95).

secondary_coverage

Secondary coverage levels for susie_rss analysis (default: c(0.7, 0.5)).

signal_cutoff

Signal cutoff for susie_post_processor (default: 0.1).

min_abs_corr

Minimum absolute correlation for credible set purity filtering (default: 0.8).

Value

A list containing the results of the SuSiE RSS analysis based on the specified method.

Details

The `susie_rss_pipeline` function runs the SuSiE RSS pipeline based on the specified analysis method. It takes the following main inputs: - `sumstats`: A list or data frame containing summary statistics with 'z' or 'beta' and 'se' columns. - `LD_mat`: The LD matrix. - `n`: Sample size (optional). - `var_y`: Variance of Y (optional). - `L`: Initial number of causal configurations to consider in the analysis. - `max_L`: Maximum number of causal configurations to consider in the analysis. - `l_step`: Step size for increasing L when the limit is reached during dynamic adjustment. - `analysis_method`: The analysis method to use. Options are "susie_rss", "single_effect", or "bayesian_conditional_regression". The function first checks if the `sumstats` input contains 'z' or 'beta' and 'se' columns. If 'z' is present, it is used directly. If 'beta' and 'se' are present, 'z' is calculated as 'beta' divided by 'se'.

Based on the specified `analysis_method`, the function calls the `susie_rss_wrapper` with the appropriate parameters. - For "single_effect" method, `L` is set to 1. - For "susie_rss" and "bayesian_conditional_regression" methods, `L`, `max_L`, and `l_step` are used. - For "bayesian_conditional_regression" method, `max_iter` is set to 1.

The results are then post-processed using the `susie_post_processor` function with the specified `signal_cutoff` and `secondary_coverage` values.

The function returns a list containing the results of the SuSiE RSS analysis based on the specified method.