tune_lucid: Parallel Grid Search for Tuning Parameters in Latent Cluster...

In USCbiostats/LUCid: Latent Unknown Clustering with Integrated Data

Description

tune_lucid fits regularized latent cluster models with various combinations of three tuning parameters based on joint inference across data types to perform a grid-search helping determine an optimal choice of three tuning parameters with minimum model BIC.

Usage

tune_lucid(
  G = NULL,
  CoG = NULL,
  Z = NULL,
  CoY = NULL,
  Y,
  K,
  Family,
  USEY = TRUE,
  initial = def_initial(),
  LRho_g,
  URho_g,
  NoRho_g,
  LRho_z_invcov,
  URho_z_invcov,
  NoRho_z_invcov,
  LRho_z_covmu,
  URho_z_covmu,
  NoRho_z_covmu,
  NoCores = detectCores() - 1
)

Arguments

G	Genetic features, a matrix
CoG	Covariates to be added in G->X path
Z	Biomarker data, a matrix
CoY	Covariates to be added in X->Y path
Y	Disease outcome, a vector
K	Pre-specified # of latent clusters
Family	"binary" or "normal" for Y
USEY	Using Y or not, default is TRUE
initial	A list of initial model parameters will be returned for integrative clustering
LRho_g	Lower limit of the penalty for selection on genetic data
URho_g	Upper limit of the penalty for selection on genetic data
NoRho_g	Number of Rho_g for grid-search
LRho_z_invcov	Lower limit of the penalty for the inverse of covariance of biomarkers
URho_z_invcov	Upper limit of the penalty for the inverse of covariance of biomarkers
NoRho_z_invcov	Number of Rho_z_invcov for grid-search
LRho_z_covmu	Lower limit of the penalty for the product of covariance and mean of biomarkers
URho_z_covmu	Upper limit of the penalty for the product of covariance and mean of biomarkers
NoRho_z_covmu	Number of Rho_z_covmu for grid-search
NoCores	Number of CPU cores for parallel grid-search, default is total number of cores minus 1

Value

tune_lucid returns an object of list containing Modelfits, Results, and Optimal:

Modelfits	Latent cluster model fits for a combination of given tuning parameters
Results	Summary results of grid-search
Optimal	Features of the optimal model with minimum BIC in the grid-search summary

Author(s)

Cheng Peng, Zhao Yang, David V. Conti

References

Cheng Peng, Jun Wang, Isaac Asante, Stan Louie, Ran Jin, Lida Chatzi, Graham Casey, Duncan C Thomas, David V Conti, A Latent Unknown Clustering Integrating Multi-Omics Data (LUCID) with Phenotypic Traits, Bioinformatics, , btz667, https://doi.org/10.1093/bioinformatics/btz667.

Examples

# For a testing dataset with 10 genetic features (5 causal) and 4 biomarkers (2 causal)
# Parallel grid-search with 8 combinations of tuning parameters
## Not run: 
GridSearch <- tune_lucid(G=G1, Z=Z1, Y=Y1, K=2, Family="binary", USEY = TRUE, NoCores = 2,
                         LRho_g = 0.008, URho_g = 0.012, NoRho_g = 2,
                         LRho_z_invcov = 0.04, URho_z_invcov = 0.06, NoRho_z_invcov = 2,
                         LRho_z_covmu = 90, URho_z_covmu = 100, NoRho_z_covmu = 2)
GridSearch$Results
# Determine the best tuning parameters
GridSearch$Optimal

## End(Not run)

USCbiostats/LUCid documentation built on Feb. 22, 2020, 8:57 p.m.