Native Bayesian Workflows

Use the native Bayesian dispatcher when you want one owned posterior run through run_bayesian_inference(...) and you are prepared to define the model surface explicitly.

This public surface is Python-first today. The stable contract is one dispatcher plus validated model-definition and proposal-schedule builders. The dispatcher rejects unsupported or mismatched inputs explicitly instead of quietly routing through external engines.

Run One Fixed-Topology DNA Chain

Fixed-topology DNA workflows keep the supplied tree constant while sampling branch lengths and supported substitution parameters.

from bijux_phylogenetics.bayesian import (
    build_dirichlet_simplex_substitution_parameter_prior,
    build_exponential_branch_length_prior,
    build_exponential_positive_substitution_parameter_prior,
    build_fixed_topology_dna_model_definition,
    build_fixed_topology_dna_proposal_schedule,
    build_substitution_parameter_prior_bundle,
    run_bayesian_inference,
)
from bijux_phylogenetics.io.fasta.core import load_fasta_alignment
from bijux_phylogenetics.io.trees import load_tree

tree = load_tree("dataset/tree.nwk")
records = load_fasta_alignment("dataset/alignment.fasta")

model_definition = build_fixed_topology_dna_model_definition(
    substitution_model_name="HKY85",
    branch_length_prior=build_exponential_branch_length_prior(rate=4.0),
    substitution_parameter_prior_bundle=build_substitution_parameter_prior_bundle(
        kappa_prior=build_exponential_positive_substitution_parameter_prior(
            rate=1.5
        ),
        base_frequency_prior=build_dirichlet_simplex_substitution_parameter_prior(
            expected_component_names=("A", "C", "G", "T"),
            concentration_parameters={"A": 2.0, "C": 2.0, "G": 2.0, "T": 2.0},
        ),
    ),
)

proposal_schedule = build_fixed_topology_dna_proposal_schedule(
    model_definition=model_definition,
    branch_length_move_weight=1.0,
    branch_length_log_scale_standard_deviation=0.25,
    kappa_move_weight=1.0,
    kappa_log_scale_standard_deviation=0.35,
    base_frequency_move_weight=1.0,
    base_frequency_coordinate_standard_deviation=0.45,
)

report = run_bayesian_inference(
    tree=tree,
    records=records,
    model_definition=model_definition,
    proposal_schedule=proposal_schedule,
    iteration_count=2_000,
    sample_every=10,
    seed=11,
)

print(report.chain_report.accepted_count)
print(len(report.posterior_rows))

The supported fixed-topology DNA substitution families are the public parameterized models, not JC69. Use K80, HKY85, or GTR when building the model definition.

Run One Joint-Topology DNA Chain

Joint-topology DNA workflows combine the same sequence-model surface with one topology prior and explicit topology-move weights.

from bijux_phylogenetics.bayesian import (
    build_exponential_branch_length_prior,
    build_exponential_positive_substitution_parameter_prior,
    build_fixed_topology_dna_model_definition,
    build_fixed_topology_dna_proposal_schedule,
    build_joint_topology_dna_model_definition,
    build_joint_topology_dna_proposal_schedule,
    build_substitution_parameter_prior_bundle,
    build_uniform_rooted_tree_topology_prior,
    run_bayesian_inference,
)
from bijux_phylogenetics.io.fasta.core import load_fasta_alignment
from bijux_phylogenetics.io.trees import load_tree

tree = load_tree("dataset/start-tree.nwk")
records = load_fasta_alignment("dataset/alignment.fasta")

sequence_model_definition = build_fixed_topology_dna_model_definition(
    substitution_model_name="K80",
    branch_length_prior=build_exponential_branch_length_prior(rate=4.0),
    substitution_parameter_prior_bundle=build_substitution_parameter_prior_bundle(
        kappa_prior=build_exponential_positive_substitution_parameter_prior(
            rate=1.5
        )
    ),
)

model_definition = build_joint_topology_dna_model_definition(
    sequence_model_definition=sequence_model_definition,
    topology_prior=build_uniform_rooted_tree_topology_prior(["A", "B", "C", "D"]),
)

proposal_schedule = build_joint_topology_dna_proposal_schedule(
    sequence_proposal_schedule=build_fixed_topology_dna_proposal_schedule(
        model_definition=sequence_model_definition,
        branch_length_move_weight=1.0,
        branch_length_log_scale_standard_deviation=0.25,
        kappa_move_weight=1.0,
        kappa_log_scale_standard_deviation=0.35,
    ),
    nni_move_weight=1.0,
)

report = run_bayesian_inference(
    tree=tree,
    records=records,
    model_definition=model_definition,
    proposal_schedule=proposal_schedule,
    iteration_count=2_000,
    sample_every=10,
    seed=7,
)

print(report.distinct_topology_count)
print(len(report.posterior_rows))

Use this surface when topology uncertainty is part of the review question. Keep the topology prior and topology-move weights explicit, because they are part of the public contract rather than hidden defaults.

What The Dispatcher Protects

The dispatcher validates that:

• the model definition is one supported native Bayesian definition
• the proposal schedule is one supported native Bayesian schedule
• fixed-topology definitions only pair with fixed-topology schedules
• joint-topology definitions only pair with joint-topology schedules

That validation is part of the serious public story. Mismatches fail explicitly instead of producing one ambiguous runtime result.

How To Read The Run Reports

Treat the returned run report as the stable output contract for the chosen workflow family.

The most important fields to inspect are:

• model_definition
• proposal_schedule
• observation_policy
• chain_report
• posterior_rows

Joint-topology runs also keep topology diversity explicit through distinct_topology_count and distinct_topology_ids.

Current Boundaries

• this public page covers the native dispatcher for fixed-topology and joint-topology DNA workflows
• this page does not claim that every Bayesian submodule has the same public maturity as those two dispatcher-routed contracts
• this page does not rename BEAST or MrBayes wrapper surfaces as native inference ownership
• this page does not replace the evidence-book or broader validation studies when the question is scientific trust rather than API usage