PoreMind documentation

PoreMind is a Python framework for single-molecule nanopore signal analysis. It converts ABF and CSV current recordings into standardized event-level data and provides signal preprocessing, event detection, feature analysis, machine learning, waveform deep learning, prediction, and visualization through a shared workflow.

Start here

• Install PoreMind and its optional readers or model dependencies.
• Follow Quick Start (API).
• Review the analysis workflow, supported models, or function reference.
• Launch the local Web UI for interactive analysis.

Key features

• A shared event representation. Detected events retain waveform data, temporal boundaries, local baseline, sample metadata, and event features so downstream analysis uses the same event records.
• Several signal-processing choices. Choose denoising, baseline, detection, and event-quality methods to fit the signal. PCA, t-SNE, and UMAP support exploration of event-feature tables.
• Classical ML and waveform DL. Train classical models on event features or use 1D-CNN, residual-CNN, and MAGJAM directly on event waveforms. MAGJAM has d8 and d4 variants.
• Three ways to work. Use the Python API, local Web UI, or the PoreMind Agent Skill.

Analysis in brief

ABF / CSV traces → preprocessing → event detection → event features and QC
                → visualization → ML / DL → prediction for new samples

This book documents the implementation in the current repository. The method framework summarizes its data model, supported methods, and current scope.