fast_HMM: Fits a HMM to a genotype dataset by calling fastPHASE
In epiGWAS: Robust Methods for Epistasis Detection

Description Usage Arguments Details Value References Examples

In this function, we fit the fastPHASE hidden Markov model (HMM) using the EM algorithm. The fastPHASE executable is required to run fast_HMM. It can be downloaded from the following web page: http://scheet.org/software.html

1 2	fast_HMM(X, out_path = NULL, X_filename = NULL, fp_path = "bin/fastPHASE", n_state = 12, n_iter = 25)

`X`	genotype matrix
`out_path`	prefix for the fitted parameters filenames. If `NULL`, the files are saved in a temporary directory.
`X_filename`	filename for the fastPHASE-formatted genotype file. If `NULL`, the file is created in a temporary directory.
`fp_path`	path to the fastPHASE executable
`n_state`	dimensionality of the latent space
`n_iter`	number of iterations for the EM algorithm

Because of the quadratic complexity of the forward algorithm in terms of the dimensionality of the latent space n_state, we recommend setting this parameter to 12. Choosing a higher number does not result in a dramatic increase of performance. An optimal choice for the number of iterations for the EM algorithm is between 20 and 25.

Fitted parameters of the fastPHASE HMM. They are grouped in a list with the following fields: pInit for the initial marginal distribution, the three-dimensional array Q for the transition probabilities and finally pEmit, another three-dimensional array for the emission probabilities

Scheet, P., & Stephens, M. (2006). A fast and flexible statistical model for large-scale population genotype data: applications to inferring missing genotypes and haplotypic phase. American Journal of Human Genetics, 78(4), 629–644.

p <- 50
n <- 100
genotypes <- matrix((runif(n * p, min = 0, max = 1) < 0.5) +
            (runif(n * p, min = 0, max = 1) < 0.5),
            nrow = n, dimnames = list(NULL, paste0("SNP_", seq_len(p))))

hmm <- fast_HMM(genotypes, fp_path = "/path/to/fastPHASE",
                n_state = 4, n_iter = 10)