65: PyPatch: Increasing reproducibility and trustworthiness of pre-clinical data

Topic: Basic Science
Subtopic: Non-specific to any diagnosis (including studies in healthy animals), Developing methodology/technology, Electrophysiology, Artificial Intelligence/Machine Learning

Background and Aims: The patch-clamp technique is a pivotal tool for the identification of promising analgesics, especially in the advent of high-throughput systems [1,3]. However, the analysis of patch-clamp data traditionally relies heavily on the manual guidance of the experimenter, which can introduce biases and may be infeasible for large datasets. Biases and changing analysis methods can hinder the reproducibility and assessment of pre-clinical findings [2]. Existing toolboxes fail to capture the decision-making process crucial to a successful analysis. We developed PyPatch as a toolbox that augments and automates the traditional analysis process and includes decisions as part of the data’s context.

Methods: We developed PyPatch as a Python based framework for the analysis of manual and automated patch-clamp data. PyPatch automates the analysis of most commonly performed patch-clamp protocols in both voltage- and current-clamp settings. However, it also offers the ability to review the data at any step of the analysis and intervene if necessary. Based on the difference between automated and manual decisions made during the analysis, explainable decision tree models can be trained. The trained models can be used to define quality control rules and be exchanged with external partners, allowing simulated decision-making on unseen data.

Results: Analysis of example datasets has shown that a large number of decisions can be taken by the automated systems implemented in the PyPatch toolbox. These rules were defined a priori based on existing literature or conventions in the patch-clamp community. Additionally, the difference between automated and manual decisions can be approximated significantly above chance level by relatively simple, explainable machine learning models. More complex models allow a more accurate modelling of the decision making at the cost of explainability.

Conclusions: PyPatch provides a powerful framework to enhance traditional analyses by keeping track of data dependencies and automating time consuming steps. It also provides scientists with tools to objectify their analyses and include decisions they made into the dataset. Thus, PyPatch not only allows a more streamlined analysis of all kinds of patch-clamp data but is also the first toolbox to make experimenter expertise part of the data’s context. We expect this added context to contribute to an overall better understanding and accessibility of data shared between labs and thus improve the reproducibility of the data.

References: [1] Bell, D.C. and Dallas, M.L. (2018) ‘Using automated patch clamp electrophysiology platforms in pain‐related ion channel research: insights from industry and academia’, British Journal of Pharmacology, 175(12), pp. 2312–2321. Available at: https://doi.org/10.1111/bph.13916.
[2] Motulsky, H.J. (2014) ‘Common Misconceptions about Data Analysis and Statistics’, The Journal of Pharmacology and Experimental Therapeutics, 351(1), pp. 200–205. Available at: https://doi.org/10.1124/jpet.114.219170.
[3] Osteen, J.D. et al. (2025) ‘Pharmacology and Mechanism of Action of Suzetrigine, a Potent and Selective NaV1.8 Pain Signal Inhibitor for the Treatment of Moderate to Severe Pain’, Pain and Therapy, 14(2), pp. 655–674. Available at: https://doi.org/10.1007/s40122-024-00697-0.

Presenting Author: Lennart Müller
Submitting Author: Lennart Müller

Senior Author: Angelika Lampert

IASP 2026 Global Year theme of Neuropathic Pain: Yes

• Marie Liselott Mehlfeldt: NA - Uniklinik RWTH Aachen
• Fiona Roll: M.Sc. - Uniklinik RWTH Aachen
• Jenny Tigerholm: Jun.-Prof. Dr. - Uniklinik RWTH Aachen
• Angelika Lampert: Prof. Dr. med. - Uniklinik RWTH Aachen