Simulated methylation QTL example

Simulated cis-mQTL data with n = 100 samples, p = 12 SNPs, and T = 32 CpGs. Three causal SNPs (1, 9, 3) act on two CpG clusters: SNPs 1 and 9 affect cluster A (CpGs 9 to 16) with opposite signs, and SNP 3 affects cluster B (CpGs 25 to 32). SNP 4 is a high-LD near-clone of SNP 3.

Usage

dnam_example

Format

A list with components

X

n x p genotype matrix (p = 12).

Y

n x T methylation matrix (T = 32).

pos

length-T integer vector of CpG indices.

causal_snps

integer vector c(1, 9, 3) of column indices in X that carry true effects.

causal_betas

length(causal_snps) x T matrix of true per-CpG effects.

truth_mask

length-3 list of length-T boolean vectors, one per CS, marking the CpGs each CS truly affects. Read by mfsusie_plot_lfsr() when the truth argument is supplied.

description

free-text description of the dataset.

Source

Simulated. See data-raw/make_data.R.

Examples

# \donttest{
data(dnam_example)
fit <- fsusie(dnam_example$Y, dnam_example$X,
              pos = dnam_example$pos, L = 15, L_greedy = 5, verbose = TRUE)
#> iter          ELBO       delta   sigma2      mem      V  extras
#>    1    -4503.3135           -   [0.988, 0.991, 0.997]   0.18 GB  [1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00]  pi_null=[1.00, 1.00]
#> iter   2: max|dPIP|=5.91e-01, V=[1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00] [mem: 0.18 GB]
#> iter   3: max|dPIP|=3.26e-01, V=[1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00] [mem: 0.18 GB]
#> iter   4: max|dPIP|=5.75e-03, V=[1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00] [mem: 0.18 GB]
#> iter   5: max|dPIP|=9.71e-03, V=[1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00] [mem: 0.18 GB]
#> iter   6: max|dPIP|=1.35e-05, V=[1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00, 1.00e+00] -- converged [mem: 0.18 GB]
#> [L_greedy] 1 round, greedy_lbf_cutoff=0.100, final L=5
#> round  L     min(lbf)   action
#> 1      5     0.000      saturated
fit_s <- mf_post_smooth(fit, method = "TI")
mfsusie_plot(fit_s)

# }