susieR

https://stephenslab.github.io/susieR
Admin
susieR is an R package that implements the Sum of Single Effects (SuSiE) model for sparse variable...
Tokens:76,877
Snippets:696
Trust Score:9.7
Update:5 months ago
Show doc for...
Tokens:
Raw
### Install susieR Development Version from GitHub

Source: https://stephenslab.github.io/susieR/index.html

Install the latest development version of susieR directly from GitHub. Ensure you have the 'remotes' package installed first.

```r
# install.packages("remotes")
remotes::install_github("stephenslab/susieR")
```

--------------------------------

### Example: Bayesian Multiple Regression with mr.ash.rss

Source: https://stephenslab.github.io/susieR/reference/mr.ash.rss.html

This example demonstrates how to generate synthetic data and then use the mr.ash.rss function to perform Bayesian multiple regression. It includes setting up prior parameters and calculating necessary summary statistics before running the main function.

```R
# Generate example data
set.seed(985115)
n <- 350
p <- 16
sigmasq_error <- 0.5
zeroes <- rbinom(p, 1, 0.6)
beta.true <- rnorm(p, 1, sd = 4)
beta.true[zeroes] <- 0

X <- cbind(matrix(rnorm(n * p), nrow = n))
X <- scale(X, center = TRUE, scale = FALSE)
y <- X %*% matrix(beta.true, ncol = 1) + rnorm(n, 0, sqrt(sigmasq_error))
y <- scale(y, center = TRUE, scale = FALSE)

# Set the prior
K <- 9
sigma0 <- c(0.001, .1, .5, 1, 5, 10, 20, 30, .005)
omega0 <- rep(1 / K, K)

# Calculate summary statistics
b.hat <- sapply(1:p, function(j) {
  summary(lm(y ~ X[, j]))$coefficients[-1, 1]
})
s.hat <- sapply(1:p, function(j) {
  summary(lm(y ~ X[, j]))$coefficients[-1, 2]
})
R.hat <- cor(X)
var_y <- var(y)
sigmasq_init <- 1.5

# Run mr.ash.rss
out <- mr.ash.rss(b.hat, s.hat,
  R = R.hat, var_y = var_y, n = n,
  sigma2_e = sigmasq_init, s0 = sigma0, w0 = omega0,
  mu1_init = rep(0, ncol(X)), tol = 1e-8, max_iter = 1e5,
  update_w0 = TRUE, update_sigma = TRUE, compute_ELBO = TRUE,
  standardize = FALSE
)



```

--------------------------------

### Example Usage of compute_marginal_bhat_shat

Source: https://stephenslab.github.io/susieR/reference/compute_marginal_bhat_shat.html

Demonstrates how to use `compute_marginal_bhat_shat` with sample matrices X and Y. It shows how to check the dimensions of the output Bhat and Shat matrices.

```R
set.seed(1)
X <- matrix(rnorm(50 * 5), 50, 5)
X <- scale(X, center = TRUE, scale = FALSE)
Y <- matrix(rnorm(50 * 3), 50, 3)
out <- compute_marginal_bhat_shat(X, Y)
dim(out$Bhat)   # 5 x 3
# > [1] 5 3
dim(out$Shat)   # 5 x 3
# > [1] 5 3
```

--------------------------------

### Example Usage of kriging_rss

Source: https://stephenslab.github.io/susieR/reference/kriging_rss.html

This example demonstrates how to use kriging_rss with simulated data. It first generates genetic data, performs univariate regression to get z-scores, calculates the correlation matrix R, and then calls kriging_rss to obtain diagnostic plots and tables. The plot generated by kriging_rss is then displayed.

```R
set.seed(1)
n <- 500
p <- 1000
beta <- rep(0, p)
beta[1:4] <- 0.01
X <- matrix(rnorm(n * p), nrow = n, ncol = p)
X <- scale(X, center = TRUE, scale = TRUE)
y <- drop(X %*% beta + rnorm(n))
ss <- univariate_regression(X, y)
R <- cor(X)
attr(R, "eigen") <- eigen(R, symmetric = TRUE)
zhat <- with(ss, betahat / sebetahat)
cond_dist <- kriging_rss(zhat, R, n = n)
cond_dist$plot
```

--------------------------------

### rss_mismatch_example

Source: https://stephenslab.github.io/susieR/reference/index.html

Real-data SuSiE-RSS example with R-reference mismatch.

```APIDOC
## rss_mismatch_example

### Description
Real-data SuSiE-RSS example with R-reference mismatch.

### Type
Data
```

--------------------------------

### Example: Fitting and plotting results with susie_auto

Source: https://stephenslab.github.io/susieR/reference/susie_auto.html

Demonstrates how to use susie_auto to fit a model and visualize the results by plotting the true coefficients against the estimated coefficients and the observed responses against the predicted responses.

```R
set.seed(1)
n = 1000
p = 1000
beta = rep(0,p)
beta[1:4] = 1
X = matrix(rnorm(n*p),nrow = n,ncol = p)
X = scale(X,center = TRUE,scale = TRUE)
y = drop(X %*% beta + rnorm(n))
res = susie_auto(X,y)
plot(beta,coef(res)[-1])
abline(a = 0,b = 1,col = "skyblue",lty = "dashed")

plot(y,predict(res))
abline(a = 0,b = 1,col = "skyblue",lty = "dashed")
```

--------------------------------

### Load Example Data for SuSiE RSS

Source: https://stephenslab.github.io/susieR/articles/rss_mismatch.html

Loads the 'rss_mismatch_example' dataset, which contains summary statistics (z, bhat, shat), sample size (n), and an LD matrix (R) for a subset of variants. This data is used to demonstrate LD mismatch scenarios.

```R
data(rss_mismatch_example, package = "susieR")
if (!exists("rss_mismatch_example"))
  load(file.path("..", "data", "rss_mismatch_example.RData"))
attach(rss_mismatch_example)
str(rss_mismatch_example)
```

--------------------------------

### Example: Trend filtering with susie_trendfilter

Source: https://stephenslab.github.io/susieR/reference/susie_trendfilter.html

Demonstrates fitting a trend filtering model to simulated data, plotting the original data, true mean, and the fitted trend. Also shows how to calculate and plot credible sets for changepoints.

```R
set.seed(1)
mu <- c(rep(0, 50), rep(1, 50), rep(3, 50), rep(-2, 50), rep(0, 200))
y <- mu + rnorm(400)
s <- susie_trendfilter(y)
#> WARNING: IBSS algorithm did not converge in 100 iterations!
plot(y)
lines(mu, col = 1, lwd = 3)
lines(predict(s), col = 2, lwd = 2)
```

```R
# Calculate credible sets (indices of y that occur just before
# changepoints).
susie_get_cs(s)
#> WARNING: Neither X nor Xcorr was provided; purity filtering is skipped and min_abs_corr will have no effect. Pass X or Xcorr to enable the purity filter.
#> $cs
#> $cs$L1
#> [1] 150
#> 
#> $cs$L2
#>  [1] 45 46 47 48 49 50 51 52 53 54 55
#> 
#> $cs$L4
#> [1] 199 200 201
#> 
#> $cs$L6
#> [1]  99 100 101 102
#> 
#> 
#> $coverage
#> [1] 0.9882658 0.9608981 0.9866797 0.9584089
#> 
#> $requested_coverage
#> [1] 0.95
#> 

# Plot with credible sets for changepoints.
susie_plot_changepoint(s, y)
#> WARNING: Neither X nor Xcorr was provided; purity filtering is skipped and min_abs_corr will have no effect. Pass X or Xcorr to enable the purity filter.
```

--------------------------------

### Simulate Data and Apply susie_trendfilter (Sparse Steps)

Source: https://stephenslab.github.io/susieR/articles/trend_filtering.html

Simulates data with a step-function mean and applies `susie_trendfilter`. This example is suitable when the underlying trend is truly sparse.

```R
library(susieR)
set.seed(1)
n=1000
mu = c(rep(0,100),rep(1,100),rep(3,100),rep(-2,100),rep(0,600))
y = mu + rnorm(n)
s = susie_trendfilter(y, 0, L=10)
# WARNING: IBSS algorithm did not converge in 100 iterations!
# WARNING: IBSS algorithm did not converge in 100 iterations!
# Warning: IBSS algorithm did not converge in 100 iterations!
```

--------------------------------

### Install susieR from CRAN

Source: https://stephenslab.github.io/susieR/index.html

Use this command to install the stable version of the susieR package from the Comprehensive R Archive Network (CRAN).

```r
install.packages("susieR")
```

--------------------------------

### Generate Synthetic Data and Fit Mr.ASH Model

Source: https://stephenslab.github.io/susieR/reference/coef.mr.ash.html

This example demonstrates generating synthetic data for a regression problem and then fitting a Mr.ASH model to this data. Ensure necessary libraries are loaded before running.

```R
set.seed(1)
n = 200
p = 300
X = matrix(rnorm(n*p),n,p)
beta = double(p)
beta[1:10] = 1:10
y = X 
fit.mr.ash = mr.ash(X, y)
```

--------------------------------

### Load susieR library

Source: https://stephenslab.github.io/susieR/articles/small_sample.html

Loads the susieR package. Ensure the package is installed before running.

```R
library(susieR)
```

--------------------------------

### Load R-reference mismatch example data

Source: https://stephenslab.github.io/susieR/reference/rss_mismatch_example.html

Loads the `rss_mismatch_example` dataset, which includes z-scores, estimated marginal regression coefficients, their standard errors, sample size estimates, and a reference correlation matrix. This data is used to test SuSiE-RSS options related to reference correlation matrices.

```R
data(rss_mismatch_example)
```

--------------------------------

### Fit SuSiE RSS with Reference Panel LD

Source: https://stephenslab.github.io/susieR/articles/susierss_diagnostic.html

Fits the SuSiE RSS model using summary statistics and a reference LD matrix. This example demonstrates how SuSiE-RSS can identify true associations even when there are allele mismatches between the summary statistics and the reference panel.

```R
fit = susie_rss(zflip, ld, n=n)
```

--------------------------------

### Get Objective Function Value in SuSiE

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Retrieves the final objective function value of a SuSiE fit. Use `last_only = FALSE` to get the history of objective function values across iterations.

```r
set.seed(1)
n <- 1000
p <- 1000
beta <- rep(0, p)
beta[1:4] <- 1
X <- matrix(rnorm(n * p), nrow = n, ncol = p)
X <- scale(X, center = TRUE, scale = TRUE)
y <- drop(X %*% beta + rnorm(n))
s <- susie(X, y, L = 10)
susie_get_objective(s)
```

```r
susie_get_objective(s, last_only = FALSE)
```

--------------------------------

### summary(<susie>) / print(<summary.susie>)

Source: https://stephenslab.github.io/susieR/reference/index.html

Summarize Susie Fit.

```APIDOC
## summary(<susie>) / print(<summary.susie>)

### Description
Summarize Susie Fit.

### Function Signature
summary(_<susie>_)
print(_<summary.susie>_)
```

--------------------------------

### Build Package Website with pkgdown

Source: https://stephenslab.github.io/susieR/index.html

This command uses the Makefile to build the package website using the pkgdown package.

```bash
make pkgdown
```

--------------------------------

### print(<summary.susie_post_outcome_configuration>)

Source: https://stephenslab.github.io/susieR/reference/index.html

Print a summary.susie_post_outcome_configuration object.

```APIDOC
## print(<summary.susie_post_outcome_configuration>)

### Description
Print a summary.susie_post_outcome_configuration object.

### Function Signature
print(_<summary.susie_post_outcome_configuration>_)
```

--------------------------------

### Get Residual Variance

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Retrieves the estimated residual variance from a SuSiE fit.

```R
susie_get_residual_variance(res)
```

--------------------------------

### Get Prior Variance

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Retrieves the estimated prior variance from a SuSiE fit.

```R
susie_get_prior_variance(res)
```

--------------------------------

### Load Libraries and Data for Fine-Mapping

Source: https://stephenslab.github.io/susieR/articles/susierss_diagnostic.html

Loads necessary libraries and simulated data for demonstrating fine-mapping diagnostics. This includes setting up genotype data, true coefficients, and calculating summary statistics.

```R
library(susieR)
library(curl)
# Using libcurl 8.20.0 with OpenSSL/3.6.2
```

```R
data("N3finemapping")
n = nrow(N3finemapping$X)
b = N3finemapping$true_coef[,1]
sumstats <- univariate_regression(N3finemapping$X, N3finemapping$Y[,1])
z_scores <- sumstats$betahat / sumstats$sebetahat
Rin = cor(N3finemapping$X)
attr(Rin, "eigen") = eigen(Rin, symmetric = TRUE)
```

--------------------------------

### get_cs_correlation()

Source: https://stephenslab.github.io/susieR/reference/index.html

Get Correlations Between CSs, using Variable with Maximum PIP From Each CS.

```APIDOC
## get_cs_correlation()

### Description
Get Correlations Between CSs, using Variable with Maximum PIP From Each CS.

### Function Signature
get_cs_correlation()
```

--------------------------------

### Simulate Data and Compare SuSiE with Genlasso (Linear + Step)

Source: https://stephenslab.github.io/susieR/articles/trend_filtering.html

Simulates data with a linear trend plus a sudden step change, then compares SuSiE and genlasso fits.

```R
set.seed(1)
mu = seq(0,4,length=1000)
mu = mu + c(rep(0,500),rep(4,500))
y = mu + rnorm(n)
s = susie_trendfilter(y,0,L=10)
y.tf = trendfilter(y,ord=0)
y.tf.cv = cv.trendfilter(y.tf)
plot(y,pch=".")
lines(mu,col=1,lwd=3)
lines(predict(s),col=2,lwd=2)
lines(y.tf$fit[,which(y.tf$lambda==y.tf.cv$lambda.min)],col=4,lwd=2)
```

--------------------------------

### Get Number of Iterations

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Retrieves the total number of iterations performed in a SuSiE fit.

```R
susie_get_niter(res)
```

--------------------------------

### Get Posterior Samples

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Draws a specified number of samples from the posterior distribution of a SuSiE fit.

```R
susie_get_posterior_samples(susie_fit, num_samples)
```

--------------------------------

### Compare susie and susie_rss with in-sample R

Source: https://stephenslab.github.io/susieR/articles/susie_rss.html

Runs both susie and susie_rss with in-sample summary statistics and plots their coefficients to show equivalence. It's recommended to estimate residual variance when R is in-sample.

```R
res1 <- susie(X,y,L = 10)
res2 <- susie_rss(bhat = ss$betahat,shat = ss$sebetahat,R = R,n = n,
                  var_y = var(y),L = 10,estimate_residual_variance = TRUE)
# HINT: Setting max_iter = 50 for the SuSiE RSS model because slow convergence is often a sign of unstable summary-statistics fitting. To disable this message, explicitly set max_iter = 50 or another value in the susie_rss() call.
# WARNING: estimate_residual_variance = TRUE is not recommended unless R is the "in-sample" R matrix; that is, the correlation matrix obtained using the exact same data matrix X that was used for the other summary statistics. If R is in-sample, set R_finite = FALSE to silence this warning. When covariates are included in the univariate regressions that produced the summary statistics, also consider removing these effects from X before computing R.
plot(coef(res1),coef(res2),pch = 20,xlab = "susie",ylab = "susie_rss")
abline(a = 0,b = 1,col = "skyblue",lty = "dashed")
```

--------------------------------

### Run SuSiE with Summary Statistics

Source: https://stephenslab.github.io/susieR/articles/finemapping.html

Use `susie_ss` for efficient fine-mapping when only summary statistics (XtX, Xty, yty) are available. This is numerically equivalent to running `susie` directly on X and y.

```R
fitted_ss <- susie_ss(XtX = ss$XtX, Xty = ss$Xty, yty = ss$yty, n = ss$n,
                      X_colmeans = ss$X_colmeans, y_mean = ss$y_mean,
                      L = 10, estimate_residual_variance = TRUE)
```

--------------------------------

### Get Log-False Discovery Rate

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Calculates the log-false discovery rate (LFSr) from a SuSiE fit.

```R
susie_get_lfsr(res)
```

--------------------------------

### Get Residual Variance from SuSiE Fit

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Extracts the estimated residual variance from a SuSiE model fit.

```r
susie_get_residual_variance(s)
```

--------------------------------

### Fine-mapping with SuSiE-ash using Summary Statistics

Source: https://stephenslab.github.io/susieR/articles/susie_unmappable_effects.html

This snippet shows how to perform fine-mapping using SuSiE-ash with summary statistics. It requires pre-calculated XtX, Xty, and yty, along with the number of samples (n). The `unmappable_effects` parameter is set to 'ash' to enable adaptive shrinkage. The output includes convergence information and memory usage.

```R
XtX <- crossprod(X)
Xty <- crossprod(X, y)
yty <- sum(y^2)
n <- nrow(X)

t0 <- proc.time()
fit_ash_ss <- susie_ss(XtX = XtX, Xty = Xty, yty = yty, n = n,
                        L = 10, unmappable_effects = "ash", verbose = TRUE)
# HINT: For SuSiE-ash, slot_prior was not specified; using default Beta(a=1, b=2), roughly expecting ~3 of 10 slots to be active. Set slot_prior = slot_prior_betabinom(a, b) to override; for an uninformative prior use slot_prior_betabinom(1, 1).
# HINT: Switching to PIP-based convergence: unmappable effects models (inf/ash) do not have a well-defined ELBO.
# iter          ELBO       delta   sigma2      mem      V
#    1     -982.7939           -   0.4676      0.57 GB  [4.36e-02, 1.75e-02, 1.16e-02, 4.12e-03, 2.08e-03, 1.54e-03, 1.22e-03, 9.88e-04, 8.14e-04, 6.76e-04], c_hat=[1.00,1.00,1.00,0.77,0.66,0.64,0.64,0.65,0.67,0.69], lbf=[40.9,12.2,6.3,1.2,0.5,0.3,0.2,0.2,0.1,0.1], C_hat=7.7(10>0.5)
# iter   2: max|d(alpha,PIP)|=2.84e-01, V=[4.51e-02, 1.56e-02, 1.25e-02, 9.46e-03, 0 x 6], c_hat=[1.00,1.00,1.00,0.99,0.69,0.69,0.70,0.70,0.71,0.71], lbf=[55.8,14.1,10.6,4.5,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=8.2(10>0.5) [mem: 0.57 GB]
# iter   3: max|d(alpha,PIP)|=3.15e-01, V=[5.01e-02, 1.68e-02, 1.30e-02, 1.32e-02, 0 x 6], c_hat=[1.00,1.00,1.00,1.00,0.71,0.71,0.71,0.71,0.71,0.71], lbf=[61.3,15.4,11.0,8.8,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=8.3(10>0.5) [mem: 0.57 GB]
# iter   4: max|d(alpha,PIP)|=2.97e-01, V=[5.38e-02, 1.68e-02, 1.33e-02, 1.14e-02, 0 x 6], c_hat=[1.00,1.00,1.00,1.00,0.71,0.71,0.71,0.71,0.71,0.71], lbf=[66.2,15.3,11.4,6.5,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=8.3(10>0.5) [mem: 0.57 GB]
# iter   5: max|d(alpha,PIP)|=2.96e-01, V=[5.40e-02, 1.70e-02, 1.21e-02, 1.12e-02, 0 x 6], c_hat=[1.00,1.00,1.00,1.00,0.71,0.71,0.71,0.71,0.71,0.71], lbf=[66.4,15.7,10.0,6.3,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=8.3(10>0.5) [mem: 0.57 GB]
# iter   6: max|d(alpha,PIP)|=2.96e-01, V=[5.38e-02, 1.69e-02, 1.20e-02, 1.12e-02, 0 x 6], c_hat=[1.00,1.00,1.00,1.00,0.71,0.71,0.71,0.71,0.71,0.71], lbf=[66.1,15.5,9.9,6.3,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=8.3(10>0.5) [mem: 0.57 GB]
# iter   7: max|d(alpha,PIP)|=2.90e-01, V=[5.75e-02, 1.64e-02, 1.23e-02, 1.68e-03, 0 x 6], c_hat=[1.00,1.00,1.00,0.73,0.69,0.69,0.69,0.69,0.68,0.68], lbf=[71.0,14.9,10.3,0.2,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=7.8(10>0.5) [mem: 0.57 GB]
# iter   8: max|d(alpha,PIP)|=3.49e-01, V=[4.99e-02, 1.71e-02, 1.28e-02, 6.23e-04, 0 x 6], c_hat=[1.00,1.00,1.00,0.68,0.68,0.67,0.67,0.67,0.67,0.67], lbf=[61.0,15.7,10.8,0.0,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=7.7(10>0.5) [mem: 0.57 GB]
# iter   9: max|d(alpha,PIP)|=3.30e-01, V=[4.74e-02, 1.77e-02, 1.27e-02, 7.06e-04, 0 x 6], c_hat=[1.00,1.00,1.00,0.68,0.67,0.67,0.67,0.67,0.67,0.67], lbf=[57.7,16.4,10.7,0.0,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=7.7(10>0.5) [mem: 0.57 GB]
# iter  10: max|d(alpha,PIP)|=3.27e-01, V=[4.68e-02, 1.79e-02, 1.27e-02, 7.91e-04, 0 x 6], c_hat=[1.00,1.00,1.00,0.68,0.67,0.67,0.67,0.67,0.67,0.67], lbf=[57.0,16.7,10.7,0.0,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=7.7(10>0.5) [mem: 0.57 GB]
# iter  11: max|d(alpha,PIP)|=3.27e-01, V=[4.68e-02, 1.80e-02, 1.27e-02, 8.15e-04, 0 x 6], c_hat=[1.00,1.00,1.00,0.68,0.67,0.67,0.67,0.67,0.67,0.67], lbf=[56.9,16.8,10.7,0.0,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=7.7(10>0.5) [mem: 0.57 GB]
# iter  12: max|d(alpha,PIP)|=8.71e-02, V=[4.68e-02, 1.80e-02, 1.27e-02, 8.19e-04, 0 x 6], c_hat=[1.00,1.00,1.00,0.68,0.67,0.67,0.67,0.67,0.67,0.67], lbf=[56.9,16.8,10.7,0.0,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=7.7(10>0.5) [mem: 0.57 GB]
# iter  13: max|d(alpha,PIP)|=4.49e-05, V=[4.68e-02, 1.80e-02, 1.27e-02, 8.19e-04, 0 x 6], c_hat=[1.00,1.00,1.00,0.68,0.67,0.67,0.67,0.67,0.67,0.67], lbf=[56.9,16.8,10.7,0.0,0.0,0.0,0.0,0.0,0.0,0.0], C_hat=7.7(10>0.5) -- converged (alpha_pip_fixed_point) [mem: 0.57 GB]
t1 <- proc.time()
t1 - t0
#    user  system elapsed 
# 205.266   3.827  74.290
susie_plot(fit_ash_ss, y = "PIP", b = b, main = "SuSiE-ash (SS)", add_legend = TRUE)
```

--------------------------------

### Fine-mapping with susie_rss using summary statistics

Source: https://stephenslab.github.io/susieR/articles/finemapping_summary_statistics.html

Performs fine-mapping using the susie_rss function with summary statistics (betahat, sebetahat), sample size (n), correlation matrix (R), and variance of Y. It estimates the residual variance and sets the maximum number of credible sets (L) to 10.

```R
fitted_rss1 <- susie_rss(bhat = sumstats$betahat, shat = sumstats$sebetahat, n = n, R = R, var_y = var(Y[,1]), L = 10,
                         estimate_residual_variance = TRUE)
# HINT: Setting max_iter = 50 for the SuSiE RSS model because slow convergence is often a sign of unstable summary-statistics fitting. To disable this message, explicitly set max_iter = 50 or another value in the susie_rss() call.
# WARNING: estimate_residual_variance = TRUE is not recommended unless R is the "in-sample" R matrix; that is, the correlation matrix obtained using the exact same data matrix X that was used for the other summary statistics. If R is in-sample, set R_finite = FALSE to silence this warning. When covariates are included in the univariate regressions that produced the summary statistics, also consider removing these effects from X before computing R.
```

--------------------------------

### Pass Slot Prior to Susie

Source: https://stephenslab.github.io/susieR/reference/slot_prior_betabinom.html

Example of how to pass a slot prior specification to the susie function for fitting a model.

```R
# Pass to susie
# fit <- susie(X, y, slot_prior = slot_prior_betabinom())
```

--------------------------------

### Get Estimated Variance (Sparse Steps)

Source: https://stephenslab.github.io/susieR/articles/trend_filtering.html

Retrieves the estimated noise variance (sigma2) from the SuSiE trend filtering result.

```R
s$sigma2
# [1] 1.076261
```

--------------------------------

### Create Initialization Object for susieR

Source: https://stephenslab.github.io/susieR/articles/l0_initialization.html

Create an initialization object for susieR using coefficients from the L0Learn fit.

```r
set.seed(1)
s.init = susie_init_coef(effect.index, effect.beta, p)
```

--------------------------------

### Get Posterior Mean

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Extracts the posterior mean of the effects from a SuSiE fit. Filters out effects with estimated prior variance smaller than a threshold.

```R
susie_get_posterior_mean(res, prior_tol = 1e-09)
```

--------------------------------

### Simulate data for comparison

Source: https://stephenslab.github.io/susieR/articles/susie_rss.html

Generates a synthetic dataset with 200 samples and 1000 variables, where the first 4 variables influence the outcome.

```R
set.seed(1)
n <- 200
p <- 1000
beta <- rep(0,p)
beta[1:4] <- 1
X <- matrix(rnorm(n*p),nrow = n,ncol = p)
X <- scale(X,center = TRUE,scale = FALSE)
y <- drop(X %*% beta + rnorm(n))
```

--------------------------------

### Get LFSR Sequence

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Generates a Linear Feedback Shift Register (LFSR) sequence. The input `s` likely represents the initial state or parameters for the LFSR.

```python
susie_get_lfsr(s)
```

--------------------------------

### Get Prior Variance from SuSiE Fit

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Retrieves the prior variance parameters used in the SuSiE model. This indicates the variance of the prior distributions for the effect sizes.

```r
susie_get_prior_variance(s)
```

--------------------------------

### Compare SuSiE RSS PIPs with in-sample vs. out-of-sample LD

Source: https://stephenslab.github.io/susieR/articles/finemapping_summary_statistics.html

Compares the PIPs from SuSiE RSS models using in-sample LD (fitted_rss1) versus out-of-sample LD (fitted_rss3). Highlights the similarity when LD matrices are close.

```R
plot(fitted_rss1$pip, fitted_rss3$pip, ylim=c(0,1), xlab='SuSiE PIP', ylab='SuSiE-RSS PIP')
```

--------------------------------

### Get Posterior Standard Deviation

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Extracts the posterior standard deviation of the effects from a SuSiE fit. Filters out effects with estimated prior variance smaller than a threshold.

```R
susie_get_posterior_sd(res, prior_tol = 1e-09)
```

--------------------------------

### Get Posterior Standard Deviations in SuSiE

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Use `susie_get_posterior_sd` to extract the posterior standard deviations for each credible set. This is useful for understanding the uncertainty around the estimated effect sizes.

```r
susie_get_posterior_sd(s)
```

--------------------------------

### Load Data and Plot Initial SuSie Model

Source: https://stephenslab.github.io/susieR/articles/susie_refine.html

Loads phenotype and genotype data from a URL and plots the initial SuSiE model results. Requires the `curl` package for downloading data.

```R
library(susieR)
library(curl)
# Using libcurl 8.20.0 with OpenSSL 3.6.2
data_file <- tempfile(fileext = ".RData")
data_url <- paste0("https://raw.githubusercontent.com/stephenslab/susieR/",
                   "master/inst/datafiles/FinemappingConvergence1k.RData")
curl_download(data_url,data_file)
load(data_file)
b <- FinemappingConvergence$true_coef
susie_plot(FinemappingConvergence$z, y = "z", b=b)
```

--------------------------------

### Summarize Susie Fit Object

Source: https://stephenslab.github.io/susieR/reference/summary.susie.html

Use this S3 method to get a summary of a 'susie' fit object. It returns a list containing variable and credible set data frames.

```R
# S3 method for class 'susie'
summary(object, ...)
```

--------------------------------

### Order Predictors from Univariate Regression

Source: https://stephenslab.github.io/susieR/reference/univar.order.html

Generates synthetic data and then uses `univar.order` to obtain the ordering of predictors based on univariate regression coefficients. Ensure the `susieR` package is installed and loaded.

```R
set.seed(1)
n           = 200
p           = 300
X           = matrix(rnorm(n*p),n,p)
beta        = double(p)
beta[1:10]  = 1:10
y           = X %*% beta + rnorm(n)

univ.order = univar.order(X,y)
```

--------------------------------

### S3 method for summary of susie_post_outcome_configuration

Source: https://stephenslab.github.io/susieR/reference/summary.susie_post_outcome_configuration.html

This is the signature for the S3 summary method. It takes the output of `susie_post_outcome_configuration` and several optional arguments to control the summarization and display.

```R
summary(
  object,
  prob_thresh = 0.8,
  ambiguous_lower = 0.5,
  signal_only = TRUE,
  color = "auto",
  ...
)
```

--------------------------------

### Compare SuSiE with Genlasso for Linear Trend

Source: https://stephenslab.github.io/susieR/articles/trend_filtering.html

Compares the SuSiE estimate (red) with the genlasso solution (blue) for a linear trend. Requires the 'genlasso' package.

```R
# install.packages("remotes")
# remotes::install_github("glmgen/genlasso")
y.tf = trendfilter(y,ord=0)
y.tf.cv = cv.trendfilter(y.tf)
plot(y,pch=".")
lines(mu,col=1,lwd=3)
lines(predict(s),col=2,lwd=2)
lines(y.tf$fit[,which(y.tf$lambda==y.tf.cv$lambda.min)],col=4,lwd=2)
```

--------------------------------

### Get Posterior Mean from SuSiE Fit

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Calculates the posterior mean of the effect sizes for each variable in the SuSiE model. This provides an estimate of the expected effect size given the data and the model.

```r
susie_get_posterior_mean(s)
```

--------------------------------

### susie_ss()

Source: https://stephenslab.github.io/susieR/reference/index.html

SuSiE using Sufficient Statistics.

```APIDOC
## susie_ss()

### Description
SuSiE using Sufficient Statistics.

### Function Signature
susie_ss()
```

--------------------------------

### Get Objective Function Value

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Retrieves the objective function (ELBO) from a SuSiE fit. Can return values from all iterations or just the last one. Warns if the ELBO decreases beyond a specified tolerance.

```R
susie_get_objective(res, last_only = TRUE, warning_tol = 1e-06)
```

--------------------------------

### Run susieR with L0Learn Initialization

Source: https://stephenslab.github.io/susieR/articles/l0_initialization.html

Run the susie function with the custom initialization object derived from L0Learn.

```r
susieL0.fit = susie(X,y,model_init=s.init)
```

--------------------------------

### Compare Direct SuSiE and Summary Statistics SuSiE

Source: https://stephenslab.github.io/susieR/articles/finemapping.html

Verify that `susie_ss` with matched hyperparameters yields the same results as running `susie` directly on the data.

```R
fitted <- susie(X, Y[,1], L = 10)
all.equal(fitted$pip, fitted_ss$pip)
```

--------------------------------

### Get Posterior Inclusion Probability

Source: https://stephenslab.github.io/susieR/reference/susie_get_methods.html

Calculates the posterior inclusion probability (PIP) for each variable from a SuSiE fit. Can optionally prune effects not included in reported credible sets and filter based on prior tolerance.

```R
susie_get_pip(res, prune_by_cs = FALSE, prior_tol = 1e-09)
```

--------------------------------

### SummaryConsistency

Source: https://stephenslab.github.io/susieR/reference/index.html

Simulated Fine-mapping Data with LD matrix From Reference Panel.

```APIDOC
## SummaryConsistency

### Description
Simulated Fine-mapping Data with LD matrix From Reference Panel.

### Type
Data
```

--------------------------------

### Load Libraries for susieR and L0Learn

Source: https://stephenslab.github.io/susieR/articles/l0_initialization.html

Load the necessary R libraries for using susieR and L0Learn.

```r
library(susieR)
library(L0Learn)
```

--------------------------------

### Get Maximum Absolute CS Correlation

Source: https://stephenslab.github.io/susieR/reference/get_cs_correlation.html

When max = TRUE, this function returns only the maximum absolute correlation among all pairs of credible sets (CSs). This is useful for a quick assessment of the highest degree of overlap.

```R
get_cs_correlation(model, max = TRUE)
```

--------------------------------

### print.summary.susie_post_outcome_configuration

Source: https://stephenslab.github.io/susieR/reference/print.summary.susie_post_outcome_configuration.html

This S3 method for the summary.susie_post_outcome_configuration class allows for pretty-printing the tidy tables generated by summary.susie_post_outcome_configuration(). It supports optional ANSI color highlighting.

```APIDOC
## print.summary.susie_post_outcome_configuration

### Description
Pretty-prints the tidy tables built by [summary.susie_post_outcome_configuration()] with optional ANSI color highlighting. See that page for the color encoding.

### Usage
```R
print(x, ...)
```

### Arguments

*   `x` (summary.susie_post_outcome_configuration): Output of [summary.susie_post_outcome_configuration()].
*   `...` (any): Ignored.

### Value
The input invisibly.

### See Also
[summary.susie_post_outcome_configuration()]
```

--------------------------------

### Get Credible Sets with Increased Purity

Source: https://stephenslab.github.io/susieR/articles/susie_unmappable_effects.html

Retrieves credible sets from a SuSiE fit using a specified minimum absolute correlation threshold. This is useful for filtering out less reliable credible sets and increasing the purity of the results.

```R
cs_pure <- susie_get_cs(fit_susie, X = X, min_abs_corr = 0.8)
cat("Number of CSs with purity >= 0.8:", length(cs_pure$cs), "\n")
# Number of CSs with purity >= 0.8: 3
```

--------------------------------

### summary(<susie_post_outcome_configuration>)

Source: https://stephenslab.github.io/susieR/reference/index.html

Summarise a susie_post_outcome_configuration result.

```APIDOC
## summary(<susie_post_outcome_configuration>)

### Description
Summarise a susie_post_outcome_configuration result.

### Function Signature
summary(_<susie_post_outcome_configuration>_)
```

--------------------------------

### susie_auto()

Source: https://stephenslab.github.io/susieR/reference/index.html

Attempt at Automating SuSiE for Hard Problems.

```APIDOC
## susie_auto()

### Description
Attempt at Automating SuSiE for Hard Problems.

### Function Signature
susie_auto()
```

--------------------------------

### Fit Mr.ASH Model and Predict Outcomes

Source: https://stephenslab.github.io/susieR/reference/predict.mr.ash.html

This snippet demonstrates generating synthetic data, fitting a Mr.ASH model, and then using the predict function to get outcomes for new data. Ensure the `mr.ash` function is available and the data dimensions are correct.

```R
set.seed(1)
n           = 200
p           = 300
X           = matrix(rnorm(n*p),n,p)
beta        = double(p)
beta[1:10]  = 1:10
y           = X %*% beta + rnorm(n)

## fit mr.ash
fit.mr.ash  = mr.ash(X, y)
#> Mr.ASH terminated at iteration 25: max|beta|=1.0032e+01, sigma2=3.0601e+00, pi0=0.7703

## predict
Xnew        = matrix(rnorm(n*p),n,p)
ypred       = predict(fit.mr.ash, Xnew)


```

--------------------------------

### Get 90% credible sets with minimum absolute correlation

Source: https://stephenslab.github.io/susieR/articles/finemapping.html

Retrieves credible sets from the fitted SuSiE object using a specified coverage of 90% and a minimum absolute correlation threshold of 0.1. This allows for potentially smaller and purer credible sets.

```r
sets <- susie_get_cs(fitted,
                     X = X,
             coverage = 0.9,
                     min_abs_corr = 0.1)
```
Tokens:
Raw