LUCID: LUCID: latent unknown clustering integrating omics with an...

In MosaiClusteR: An Umbrella Framework for Multi-Source and Multi-Omics Clustering

LUCID: latent unknown clustering integrating omics with an outcome

Description

Thin wrapper around LUCIDus' estimate_lucid. LUCID is a model-based, supervised integrative clustering method: it defines a categorical latent variable (the clusters) jointly from upstream exposures G, multi-omics data Z and an outcome Y via an EM algorithm (a quasi-mediation model). Unlike the unsupervised methods in MosaiClusteR it requires an outcome and returns posterior cluster membership.

Usage

LUCID(
  G,
  Z,
  Y,
  K,
  lucid_model = c("early", "parallel", "serial"),
  family = c("normal", "binary"),
  ...
)

Arguments

G	Exposure/predictor matrix (objects in rows).
Z	Omics matrix, or a list of matrices for the parallel/serial models (objects in rows).
Y	Outcome vector (length = number of objects).
K	Number of latent clusters.
lucid_model	One of "early", "parallel", "serial".
family	Outcome family, "normal" or "binary".
...	Further arguments forwarded to LUCIDus::estimate_lucid.

Value

A list of class "LUCID" with the fitted model and the posterior cluster assignment per object.

References

Zhao, Y., Jia, K., Goodrich, J. & Conti, D. V. (2024). LUCIDus: an R package for implementing LUCID with phenotypic traits. The R Journal 16(2).

See Also

M_ABC, intNMF

Examples

## Not run: 
# requires the 'LUCIDus' package
data(mosaic_toy)
Z <- mosaic_toy$List[[1]]
G <- mosaic_toy$List[[2]][, 1:3]
Y <- as.numeric(mosaic_toy$truth)
fit <- LUCID(G = G, Z = Z, Y = Y, K = 3)
table(fit$cluster)

## End(Not run)

MosaiClusteR documentation built on Oct. 25, 2026, 1:08 a.m.