Analysis Pipeline Overview

The CasTuner Python port is structured as a seven-step pipeline, orchestrated via Snakemake:

  1. Kinetic fitting of up- and down-regulation half-times (Step 1a–c),
  2. Hill curve fitting for steady-state dose–response (Step 1c),
  3. ODE-based derepression and repression simulations (Steps 2–3),
  4. Single-cell noise analysis (Step 4a),
  5. Model-driven design-space scan (Step 4b),
  6. Goodness-of-fit diagnostics (Step 6),
  7. Design selection & mapping (Step 7).

A final step collects all tables and plots into a single PDF report.

High-level data flow

  1. Read and gate flow cytometry .fcs files
    All steps use consistent:

    • FSC/SSC gating,
    • singlet gating,
    • NFC-based background subtraction.

  2. Compute summary statistics per condition

    • medians for BFP/mCherry/EGFP,
    • CV² for noise analysis,
    • dose-normalised readouts.

  3. Fit phenomenological models

    • Exponential functions → half-times of BFP rise/decay,
    • Hill functions → K and n for steady-state dose–response.

  4. Fit mechanistic ODE model

    • one model for derepression,
    • one model for repression,
    • shared parameters: Hill K, n and mCherry degradation rate α,
    • inferred delays for REV and KD.

  5. Propagate uncertainty & explore design space

    • sample designs from measured parameter ranges,
    • simulate trajectories,
    • score candidates by dynamic range + speed.

  6. Summarise

    • integrated plots:
      • kinetics vs noise,
      • K vs n,
      • delays vs half-times,
    • goodness-of-fit plots (obs vs pred),
    • final design space map and top-10 candidate table.

  7. Generate automated report

    • Compile all results into a single PDF.
    • Sections for each analysis step with methods, results, interpretations.
    • Tables and plots embedded inline.

  8. Sensitivity & uncertainty analysis

    • Quantify parameter sensitivities and output uncertainties.
    • Generate additional plots and tables.

Each of the following pages explains one step in more depth: what biological question it answers, what inputs it uses, which models it fits, and how to interpret the outputs.