nunchaku

nunchaku implements a Bayesian method to partition 1D time series data into contiguous piece-wise linear segments for objective identification of discontinuous change points in quantitative biological measurements.

Key Features:

• Bayesian framework: Employs a Bayesian methodology to infer optimal partitioning of data into contiguous piece-wise linear segments.
• Support for arbitrary basis functions: Extends beyond simple linear relationships by allowing segmentation based on linear combinations of arbitrary basis functions.
• Change-point identification: Detects discontinuous change points within datasets to locate boundaries between distinct linear segments.
• Automation and high throughput: Algorithmic design supports automated analyses to enable high-throughput and reproducible segmentation workflows.
• Python implementation: Provided as a Python package for integration into computational analysis environments.
• Statistical rigor: Uses a statistically rigorous approach to minimize subjective interpretation when defining segment boundaries.

Scientific Applications:

• Microbial growth analysis: Determines the optical density range where OD is linearly proportional to cell count and identifies regions of exponential growth for organisms such as Escherichia coli and Saccharomyces cerevisiae.
• Inference of biological constants: Enables inference of parameters such as the Monod constant for budding yeast growing on fructose from segmented growth data.

Methodology:

Applies a Bayesian partitioning approach that infers optimal segmentation of datasets into segments described by linear relationships or linear combinations of basis functions.