R package with matrix factorization algorithms
85 | 86 | 87 | 88 | 89 |Class which store data
85 | 86 | 87 |Dat(Y)88 | 89 | 90 |
Fit the model
85 | 86 | 87 |lfmm_fit(m, dat, ...)88 | 89 | 90 |
return a list of train/test indices
85 | 86 | 87 |left.out.kfold(kfold, J)88 | 89 | 90 |
Impute Y with a fitted model.
85 | 86 | 87 |lfmm_impute(m, dat, ...)88 | 89 | 90 |
Compute the residual error
85 | 86 | 87 |lfmm_residual_error2(m, dat, ...)88 | 89 | 90 |
Cross validation
85 | 86 | 87 |lfmm_CV(m, dat, n.fold.row, n.fold.col, ...)88 | 89 | 90 |
Class which store data
85 | 86 | 87 |LfmmDat(Y, X, missing = TRUE)88 | 89 | 90 |
linear model: 85 | Y = X B^T + E
86 | 87 | 88 |hypothesis_testing_lm(dat, X)89 | 90 | 91 |
see mon cahier 6/07/2017
85 | 86 | 87 |compute_P(X, lambda)88 | 89 | 90 |
Class which store data
85 | 86 | 87 |SimulatedLfmmDat(Y, X, outlier, U, V, B)88 | 89 | 90 |
Fit the model when latent factor loadings are known
85 | 86 | 87 |lfmm_fit_knowing_loadings(m, dat, ...)88 | 89 | 90 |
Fit assuming V and B
85 | 86 | 87 |# S3 method for ridgeLFMM 88 | lfmm_fit_knowing_loadings(m, dat)89 | 90 | 91 |
score are assume to follow student distibution with df degre of freedom
85 | 86 | 87 |compute_pvalue_from_tscore(score, df)88 | 89 | 90 |
score are assume to follow normal distibution
85 | 86 | 87 |compute_pvalue_from_zscore(score, mean = 0, sd = 1)88 | 89 | 90 |
A dataset containing SNP frequency and simulated phenotypic data for 170 plant accessions. 85 | The variables are as follows:
86 | 87 | 88 |data(example.data)89 | 90 |
A list with 4 arguments: genotype, phenotype, causal.set, chrpos
93 | 94 |genotype: binary (0 or 1) SNP frequency for 170 individuals (26943 SNPs).
phenotype: simulated phenotypic data for 170 individuals.
causal.set: set of indices for causal SNPs.
chrpos: genetic map including chromosome position of each SNP.
A data set containing normalized beta values, and sun exposure and simulated 85 | phenotypic data for 78 tissue samples.
86 | 87 | 88 |data("skin.exposure")89 | 90 |
A list with 6 arguments: beta.value, phenotype, causal.set, chrpos
93 | 94 |The variables are:
beta.value: 1496 filtered normalized beta values (methyation probabilities) 98 | for 78 tissue samples.
exposure: Sun exposure levels for 78 tissue samples.
phenotype: Simulated binary phenotypic data for 78 tissue samples.
age: age of patients.
gender: sex of patients.
tissue: category for tissue samples.
Reference: to be filled
106 | 107 | 108 |Genome and epigenome-wide association studies are plagued with the problems of confounding and causality. The R package lfmm implements new algorithms for parameter estimation in latent factor mixed models (LFMM). The algorithms are designed for the correction of unobserved confounders. The new methods are computationally efficient, and provide statistically optimal corrections resulting in improved power and control for false discoveries. The package lfmm provides two main functions for estimating latent confounders (or factors): lfmm_ridge and lfmm_lasso. Those functions are based on optimal solutions of regularized least-squares problems. A short tutorial provides brief examples on how the R packages lfmm can be used for fitting latent factor mixed models and evaluating association between a response matrix (SNP genotype or methylation levels) and a variable of interest (phenotype or exposure levels) in genome-wide (GW), genome-environment (GE), epigenome-wide (EW) association studies. Corresponding software is available at the following url https://bcm-uga.github.io/lfmm/.
Installing the latest version from github requires devtools:
66 |# install.packages("devtools")
67 | devtools::install_github("bcm-uga/lfmm")