bletl

bletl integrates BioLector microcultivation device data into Python workflows to support microbial phenotyping and bioprocess characterization within the Design-Build-Test-Learn cycle.

Key Features:

• Data integration and parsing: Reads raw result files from BioLector I, II, and Pro and parses contained information into Python data structures.
• Spline-based derivative analysis: Performs spline-based derivative calculations on time-series signals from Microbioreactor (MBR) devices.
• Unbiased growth-rate quantification: Quantifies time-variable specific growth rate \vec{\mu}_t using unsupervised switchpoint detection based on Student-t distributed random walks with Bayesian uncertainty quantification.
• Automatic switch-point detection: Detects switch-points automatically to identify significant metabolic changes in cultivation time series.
• Time-series feature extraction: Extracts time-series features from MBR data to support downstream analytical and machine learning workflows.
• Machine learning visualization: Supports machine learning–based unsupervised phenotype characterization and visualization, including Neighbor Embedding (t-SNE) of growth/DO/pH phenotypes.
• Python scientific ecosystem compatibility: Leverages standard tools from the Python scientific computing ecosystem for analysis and visualization.

Scientific Applications:

• Microbial phenotyping: Enables quantitative characterization of microbial growth dynamics and phenotypic variation from BioLector MBR experiments.
• Bioprocess characterization: Facilitates detection and quantification of metabolic changes and bioprocess dynamics in high-throughput microcultivation data.
• Unsupervised phenotype discovery: Supports unsupervised machine learning workflows to identify and visualize phenotypic clusters across growth, DO, and pH signals.

Methodology:

Reads and parses raw BioLector result files; computes spline-based derivatives; performs unsupervised switchpoint detection using Student-t distributed random walks to quantify \vec{\mu}_t with Bayesian uncertainty quantification; extracts time-series features and applies Neighbor Embedding (t-SNE) for visualization.