Kinbiont

Kinbiont is a Julia package designed as an end-to-end pipeline for biological discovery, enabling data-driven generation of hypotheses that can be tested in targeted experiments. Leveraging Julia's speed, flexibility, and growing popularity in biological, Kinbiont integrates advanced solvers for ordinary differential equations (ODEs), non-linear optimization methods, signal processing, and interpretable ML algorithms. Kinbiont can fit various models of microbial dynamics, including solutions in closed-form, ODEs, and user-defined models. Unlike existing tools, Kinbiont extends model-based parameter estimation to fits with segmentation, allowing for more detailed analysis of the microbial dynamics. The inferred parameters can then be analyzed using symbolic regression to discover equations that capture response patterns or decision tree to identify informative response patterns from large datasets.

New to Kinbiont? Start with the Quick Start — you can go from a CSV file to fitted growth parameters in under 10 lines of Julia.

• Installation & requirements
• • Manual installation
  • Package installation
  • Requirements & Dependencies
• Quick Start
• • Track 1 — I have a CSV file
  • Track 2 — I'm building from a DataFrame
  • Track 3 — Full round-trip (simulate → save → reload → fit)
  • Next steps
• Data & Types
• • GrowthData
  • CSV format
  • Annotation format
  • FitOptions
  • ModelSpec
  • MODEL_REGISTRY
  • Result types
  • CSV format for multi-dimensional data
  • ML input format
• Preprocessing
• • Smoothing
  • Blank subtraction
  • Replicate averaging
  • Negative value correction
  • Stationary phase cutting
  • Scattering correction
  • Real-data example
• Fitting
• • The three model types
  • Discovering available models
  • Fitting a single curve
  • Model selection
  • Bounds and initial guesses
  • Fitting a whole plate
  • Multistart
  • Subsetting and re-fitting
  • Log-linear fitting
  • Legacy API
• Clustering
• • Basic usage
  • Choosing k — elbow plot
  • Handling non-growing wells
  • Exponential prototype cluster
  • Reproducibility
  • Synthetic example: comparing clustering modes
  • Inspecting and subsetting clusters
  • Two-step workflow: cluster then fit
• Clustering analysis of growth curves
• • Clustering of growth curves
  • Basic clustering on z-scored curves
  • Behaviour of row-wise z-score normalisation
  • Constant pre-screening of flat curves
  • Trend-test reassignment of flat curves
  • Priority between pre-screening and trend-test modes
  • Exponential prototype relabeling
  • Centroid computation
  • Repeated k-means initialisations
  • Edge case: empty input
  • Full clustering workflow example
  • Summary
• ML Downstream
• • Decision tree regression — real data
  • Decision tree regression — synthetic data
  • Symbolic regression — synthetic data
• Examples of Simulations
• • Simulating One-Dimensional Data with ODEs
  • Simulating Data with Stochastic Simulations
  • Simulating ODEs System
  • Simulating ODEs System with Custom Model
  • Simulating Reaction Network
  • Simulating Cybernetic Models in Kinbiont
  • Data Preprocessing
• The mathematical models
• • NL models for bacterial growth
  • ODEs for bacterial growth
  • Stochastic models for bacterial growth
  • ODE Systems
  • Cybernetic models for bacterial growth
  • Reaction networks
• API Reference
• • Pre-processing functions
  • Fitting one kinetics
  • Fitting one a .csv file
  • ML downstrema analysis
  • Simulations
  • Various