hmmprofestimate

Estimate profile hidden Markov model (HMM) parameters using pseudocounts

Syntax

hmmprofestimate(Model, MultipleAlignment,'PropertyName',PropertyValue,...)

hmmprofestimate(..., 'A', AValue)
hmmprofestimate(..., 'Ax', AxValue)
hmmprofestimate(..., 'BE', BEValue)
hmmprofestimate(..., 'BDx', BDxValue)

Arguments

Model

Hidden Markov model created with the function hmmprofstruc.

MultipleAlignment

Array of sequences. Sequences can also be a structured array with the aligned sequences in a field Aligned or Sequences, and the optional names in a field Header or Name.

A

Property to set the pseudocount weight A. Default value is 20.

Ax

Property to set the pseudocount weight Ax. Default value is 20.

BE

Property to set the background symbol emission probabilities. Default values are taken from Model.NullEmission.

BMx

Property to set the background transition probabilities from any MATCH state ([M->M M->I M->D]). Default values are taken from hmmprofstruct.

BDx

Property to set the background transition probabilities from any DELETE state ([D->M D->D]). Default values are taken from hmmprofstruct.

Description

hmmprofestimate(Model, MultipleAlignment, 'PropertyName', PropertyValue...) returns a structure with the fields containing the updated estimated parameters of a profile HMM. Symbol emission and state transition probabilities are estimated using the real counts and weighted pseudocounts obtained with the background probabilities. Default weight is A=20, the default background symbol emission for match and insert states is taken from Model.NullEmission, and the default background transition probabilities are the same as default transition probabilities returned by hmmprofstruct.

Model Construction: Multiple aligned sequences should contain uppercase letters and dashes indicating the model MATCH and DELETE states agreeing with Model.ModelLength. If model state annotation is missing, but MultipleAlignment is space aligned, then a "maximum entropy" criteria is used to select Model.ModelLength states.

    Note:   Insert and flank insert transition probabilities are not estimated, but can be modified afterwards using hmmprofstruct.

hmmprofestimate(..., 'A', AValue) sets the pseudocount weight A = Avalue when estimating the symbol emission probabilities. Default value is 20.

hmmprofestimate(...,'Ax', AxValue) sets the pseudocount weight Ax = Axvalue when estimating the transition probabilities. Default value is 20.

hmmprofestimate(...,'BE', BEValue) sets the background symbol emission probabilities. Default values are taken from Model.NullEmission.

hmmprofestimate(...,'BMx', BMxValue) sets the background transition probabilities from any MATCH state ([M->M M->I M->D]). Default values are taken from hmmprofstruct.

hmmprofestimate(..., 'BDx', BDxValue) sets the background transition probabilities from any DELETE state ([D->M D->D]). Default values are taken from hmmprofstruct.

Was this topic helpful?