infer_profiles_vb: Infer methylation profiles using VB
In BPRMeth: Model higher-order methylation profiles

Description Usage Arguments Value Details Author(s) See Also Examples

General purpose functions for inferring latent profiles for different observation models using Variational Bayes (VB). Current observation models are: 'bernoulli', 'binomial' or 'gaussian'.

infer_profiles_vb(
  X,
  model = NULL,
  basis = NULL,
  H = NULL,
  w = NULL,
  gaussian_l = 50,
  alpha_0 = 0.5,
  beta_0 = 0.1,
  vb_max_iter = 100,
  epsilon_conv = 1e-05,
  is_parallel = FALSE,
  no_cores = NULL,
  is_verbose = FALSE,
  ...
)

`X`	The input data, either a `matrix` or a `list` of elements of length N, where each element is an `L X C` matrix, where L are the total number of observations. The first column contains the input observations x (i.e. CpG locations). If "binomial" model then C=3, and 2nd and 3rd columns contain total number of trials and number of successes respectively. If "bernoulli" or "gaussian" model, then C=2 containing the output y (e.g. methylation level).
`model`	Observation model name as character string. It can be either 'bernoulli', 'binomial', 'beta' or 'gaussian'.
`basis`	A 'basis' object. E.g. see `create_basis`. If NULL, will an RBF object will be created.
`H`	Optional, design matrix of the input data X. If NULL, H will be computed inside the function.
`w`	A vector of initial parameters (i.e. coefficients of the basis functions). If NULL, it will be initialized inside the function.
`gaussian_l`	Noise precision parameter, only used when having "gaussian" observation model.
`alpha_0`	Hyperparameter: shape parameter for Gamma distribution. A Gamma distribution is used as prior for the precision parameter tau.
`beta_0`	Hyperparameter: rate parameter for Gamma distribution. A Gamma distribution is used as prior for the precision parameter tau.
`vb_max_iter`	Integer denoting the maximum number of VB iterations.
`epsilon_conv`	Numeric denoting the convergence threshold for VB.
`is_parallel`	Logical, indicating if code should be run in parallel.
`no_cores`	Number of cores to be used, default is max_no_cores - 1.
`is_verbose`	Logical, print results during VB iterations.
`...`	Additional parameters.

An object of class infer_profiles_vb_"obs_model" with the following elements:

W: An Nx(M+1) matrix with the optimized parameter values. Each row of the matrix corresponds to each element of the list X; if X is a matrix, then N = 1. The columns are of the same length as the parameter vector w (i.e. number of basis functions).
W_Sigma: A list with covariance matrices for each element row in W.
basis: The basis object.
nll_feat: NLL fit feature.
rmse_feat: RMSE fit feature.
coverage_feat: CpG coverage feature.
lb_feat: Lower Bound feature.

The modelling and mathematical details for inferring profiles using mean-field variational inference are explained here: http://rpubs.com/cakapourani/ . More specifically:

For Binomial/Bernoulli observation model check: http://rpubs.com/cakapourani/variational-bayes-bpr
For Gaussian observation model check: http://rpubs.com/cakapourani/variational-bayes-lr

C.A.Kapourani C.A.Kapourani@ed.ac.uk

create_basis, infer_profiles_mle, predict_expr, create_region_object

# Example of inferring parameters for synthetic data using 3 RBFs
basis <- create_rbf_object(M=3)
out <- infer_profiles_vb(X = binomial_data, model = "binomial",
   basis = basis, is_parallel = FALSE, vb_max_iter = 10)

#-------------------------------------

basis <- create_rbf_object(M=3)
out <- infer_profiles_vb(X = gaussian_data, model = "gaussian",
   basis = basis, is_parallel = FALSE, vb_max_iter = 10)