​
High-level description

This directory contains unit tests for branch length estimators in the Cassiopeia library, specifically focusing on two classes: IIDExponentialBayesian and IIDExponentialMLE. These tests are designed to validate the functionality and accuracy of these estimators under various scenarios, including edge cases, hand-solvable problems, and simulated data.

​
What does it do?

The tests in this directory serve several purposes:
  1. Validate the correctness of the IIDExponentialBayesian estimator by comparing its output to closed-form solutions and numerical approximations for log likelihood, log joints, and posterior time distributions.
  2. Verify the behavior of the IIDExponentialMLE estimator in various scenarios, including:
    • Degenerate cases (no mutations or saturation)
    • Hand-solvable problems with simple tree topologies
    • Regression tests on small trees
    • Performance on simulated data
    • Handling of subtree collapses when no mutations are present
    • Enforcement of minimum branch length constraints
    • Incorporation of site-specific mutation rates
  3. Ensure proper error handling for invalid inputs or edge cases in both estimators.
These tests help maintain the reliability and accuracy of the branch length estimation algorithms in the Cassiopeia library, which is crucial for phylogenetic analysis and lineage tracing applications.

​
Key Files

  1. iid_exponential_bayesian_test.py: This file contains tests for the IIDExponentialBayesian class. It includes:
    • Comparisons against closed-form solutions for small and medium-sized trees
    • Validation of error handling for invalid inputs
    • Numerical calculations of log likelihood, log joints, and posterior time distributions
  2. iid_exponential_mle_test.py: This file contains tests for the IIDExponentialMLE class. It includes:
    • Tests for degenerate cases (no mutations and saturation)
    • Verification of results for hand-solvable problems
    • Regression tests on small trees
    • Tests using simulated data
    • Validation of subtree collapse behavior
    • Tests for minimum branch length enforcement
    • Verification of site-specific mutation rate handling
Both files use parameterized testing to run the same tests with different solvers (ECOS and SCS) where applicable.

​
Dependencies

The tests rely on several external libraries and frameworks:
  1. unittest: The standard Python unit testing framework.
  2. parameterized: Used for creating parameterized tests, allowing the same test to be run with different inputs.
  3. numpy: Used for numerical operations and array manipulations.
  4. scipy: Specifically scipy.integrate is used for numerical integration in the Bayesian tests.
  5. networkx: Used for working with graph structures, particularly in creating and manipulating tree topologies.
  6. cvxpy: A Python-embedded modeling language for convex optimization problems, used in the MLE estimator.
From the Cassiopeia library:
  • cassiopeia.data.CassiopeiaTree: Represents the phylogenetic tree structure.
  • cassiopeia.simulator.Cas9LineageTracingDataSimulator: Used for generating simulated data.
  • cassiopeia.tools.IIDExponentialBayesian and cassiopeia.tools.IIDExponentialMLE: The classes being tested.

​
Configuration

The tests do not rely on external configuration files or environment variables. However, they do use various parameters to configure the estimators and test scenarios:
  1. For IIDExponentialBayesian:
    • mutation_rate: The mutation rate of the model.
    • birth_rate: The birth rate of the model.
    • sampling_probability: The sampling probability of the model.
    • discretization_level: The number of timesteps used to discretize time.
  2. For IIDExponentialMLE:
    • solver: The optimization solver to use (ECOS or SCS).
    • minimum_branch_length: A constraint on the minimum allowed branch length.
    • relative_mutation_rates: Site-specific mutation rates.
These parameters are set within the test methods to create various test scenarios and validate the estimators’ behavior under different conditions. In summary, this directory contains comprehensive unit tests for two branch length estimators in the Cassiopeia library, ensuring their accuracy and reliability across a wide range of scenarios and input conditions.