Abstract:
This thesis develops and validates a reproducible workflow for AFM cell rheology that adapts the open-source PyFMLab suite to JPK NanoWizard 2 force–indentation datasets and bench- marks it against a vetted MATLAB implementation. A central contrast is methodological: PyFMLab uses an analytical (closed-form) Ting contact solution, whereas the MATLAB workflow evaluates Ting’s model numerically on the same curves. Four predefined MEF cohorts—wild type, Dyna- sore (dynamin inhibition), Y–27632 (ROCK inhibition), and the dual treatment—are analyzed exactly as provided; wet-lab execution is out of scope, and biology is included only for naming and interpretation. Raw deflection and base-piezo signals are converted to force and indentation using a stan- dardized calibration. Curves then pass objective goodness-of-fit screens and quick visual checks before full-cycle linear viscoelastic fitting with power-law relaxation to estimate two compact parameters: the instantaneous modulus E0 and a fluidity exponent. Three candidate inclusion criteria were tested across sessions; after reviewing edge cases, a stringent rule was fixed for primary MATLAB reporting to prioritize high-reliability curves. Results are summarized at curve and cell levels using descriptive statistics, box-and-whisker plots, and omnibus nonparametric tests with corrected post hoc contrasts, enabling transparent cross-tool comparison of E0 and the exponent. Both pipelines support the same qualitative bi- ology: ROCK inhibition shows the largest increase in fluidity, the dual treatment is elevated, and Dynasore remains near WT; stiffness trends are consistent with softening under ROCK inhibition within the probed window. Rather than exact numerical parity, the side-by-side anal- ysis reveals systematic offsets attributable to implementation choices, including the solution method for Ting’s model (analytical vs. numerical), the PLR kernel form, the fitting domain (force–indentation vs. force–time), contact detection/initialization, residual weighting and win- dowing, optimizer settings, and QC retention. In practice, PyFMLab delivers markedly faster throughput and, beyond quasi-static indentation, supports in-contact dynamic microrheology, whereas the MATLAB workflow used here focuses on quasi-static cycles. The study motivates harmonized preprocessing, explicit configuration reporting, and clear QC criteria for repro- ducible, interpretable AFM viscoelastic readouts.
