methylseg.CTMethylSegHMM

class methylseg.CTMethylSegHMM(n_states, n_emissions=4, holding_time_guess=1500000, time_scale=1, max_iter=25, tol=0.01, random_state=42, algorithm='forward-backward')[source]

Bases: MethylSegHMM

Continuous-time HMM backend for sparsely spaced CpGs along a chromosome.

Parameters:

  • n_emissions (int)

  • holding_time_guess (int)

  • time_scale (float)

  • max_iter (int)

  • tol (float)

  • random_state (int)

__init__(n_states, n_emissions=4, holding_time_guess=1500000, time_scale=1, max_iter=25, tol=0.01, random_state=42, algorithm='forward-backward')[source]

Initialize a continuous-time HMM for unevenly spaced CpGs.

Parameters:

  • n_states – Number of hidden methylation states.

  • n_emissions (int) – Number of discrete observed emission categories.

  • holding_time_guess (int) – Initial genomic holding-time scale in base pairs.

  • time_scale (float) – Multiplier applied to genomic time intervals.

  • max_iter (int) – Maximum fitting iterations for the continuous-time backend.

  • tol (float) – Convergence tolerance for continuous-time fitting.

  • random_state (int) – Random seed passed to the continuous-time backend.

  • algorithm – Fitting algorithm supported by the continuous-time backend.

Methods

__init__(n_states[, n_emissions, ...])

Initialize a continuous-time HMM for unevenly spaced CpGs.

create_model()

Create the continuous-time HMM with a default near-identity emission model.

fit(emissions, sample_info, chrom)

Fit the continuous-time HMM using CpG coordinates as observation times.

format_fit(emissions)

Pair observed states with genomic coordinates for CT-HMM fitting.

format_predict(emissions)

Pair observed states with genomic coordinates for CT-HMM decoding.

predict(emissions)

Decode hidden states with the configured continuous-time algorithm.
create_model()[source]

Create the continuous-time HMM with a default near-identity emission model.

format_fit(emissions)[source]

Pair observed states with genomic coordinates for CT-HMM fitting.

Parameters:

emissions – Integer observed state sequence.

Returns:

Single-sequence list containing (observed_states, times).

Return type:

list

format_predict(emissions)[source]

Pair observed states with genomic coordinates for CT-HMM decoding.

Parameters:

emissions – Integer observed state sequence.

Returns:

(observed_states, times) for the decoder.

Return type:

tuple

fit(emissions, sample_info, chrom)[source]

Fit the continuous-time HMM using CpG coordinates as observation times.

Parameters:

  • emissions – Integer observed state sequence for one chromosome.

  • sample_info – Sample metadata whose methylation table supplies CpG genomic positions.

  • chrom – Chromosome whose CpGs should be used to derive observation times.

Returns:

Fitted CT-HMM object returned by cthmm.

Return type:

object

predict(emissions)[source]

Decode hidden states with the configured continuous-time algorithm.

Parameters:

emissions – Integer observed state sequence.

Returns:

Decoded hidden-state assignments.

Return type:

numpy.ndarray