Tag

Reproducibility.

Every dispatch tagged Reproducibility. Evidence status and limitations are stated in each article.

AI in the labSep 8, 20265 min

AI for wet-lab analysis: what to hand over

A practical guide to giving an AI assistant the data, controls, units, and replicate structure it needs, with a worked example of paired donor measurements.

Thirty-six constructed well measurements compared with six paired donor means, showing the experimental structure hidden by pooled observations

Δ vs paper

Design example

№

Inside the softwareSep 8, 20265 min

How an AI analysis agent works

Follow an agent from an uploaded spreadsheet through tool calls, calculations, questions, and reported results. A plain-language look at agentic AI for lab analysis.

Diagram connecting input inspection, model action selection, calculation, result review, researcher clarification, and reporting

Δ vs paper

Explainer

№

Analysis reviewSep 8, 20265 min

How to check an AI-generated lab analysis

Review inputs, residuals, fitted values, exports, and scientific claims using a documented enzyme-kinetics run and an independent calculation on published data.

Residuals of independent Michaelis–Menten fits to treated and untreated Puromycin observations on matching vertical scales

Δ vs paper

Evidence review

№

Flow cytometrySep 8, 20265 min

Flow cytometry gating for beginners

Follow an annotated gating hierarchy from all events to CD3+ CD4+ populations. Learn controls, singlet gates, viability, and parent percentages.

Four annotated plots of simulated events showing scatter, singlet, viability, and CD3/CD4 gates

Δ vs paper

Worked example

№

Dose-responseSep 8, 20265 min

Why your IC50 curve fit fails

Troubleshoot incomplete dose-response curves, normalization, outliers, and constraints, with a reproducible example of relative versus absolute IC50.

Simulated inhibitory curve showing a restricted concentration range and distinct relative and absolute 50-percent endpoints

Δ vs paper

Worked example

№

Standard curvesSep 8, 20264 min

4PL vs. 5PL curve fitting: how to choose

Compare 4PL and 5PL standard curves using residuals and back-calculated concentrations, with downloadable simulated data and a reproducible Python script.

Simulated asymmetric calibration standards with 4PL and 5PL fitted curves and residuals

Δ vs paper

Worked example

№

Open-data analysisSep 8, 20263 min

Checking a live enzyme-kinetics analysis

A production run on the open Puromycin dataset, checked against an independent SciPy fit. Results, reproducible inputs, and workflow limitations.

Observed Puromycin reaction rates and curves reconstructed from the live app parameters

Δ vs paper

Numerical check

All field notes Browse all tags