locpix_points, aka ClusterNet, for graph-based deep learning of supracluster structure in point-cloud data
Release v0.1.1 · oubino/locpix_points / oubino/locpix_points at v0.1.1 This software is for classification of point-cloud data based on the features and spatial arrangement of clusters within the data. It uses graph-based neural networks, taking the point coordinates and their assignment into clusters as input (k-means clustering and DBSCAN are included). It can use handcrafted geometrical cluster features or learn discriminative cluster features itself. It can identify subgraphs from the processed data that are discriminative for classification to include explainability with SubgraphX. ClusterNet was first developed for classifying data from single-molecule localisation microscopy [1], but is applicable to any point cloud data with hierarchical structure. This software was first described and demonstrated in [2]. Data for replicating the classification of digits and letters are in demo/, with instructions in README.rst and specifically "Reproducing manuscript results". Classification of nanoscale protein distribution in metastatic colorectal cancer patients to predict response to therapy was developed and is included in paper/ereg-pan-2026. Instructions for following this modelling pipeline are in paper/ereg-pan-2026/README.txt. See README.rst and addenda.rst for more information. [1] Lelek, M., et al. (2021). "Single-molecule localization microscopy." Nature Reviews Methods Primers 1(1): 39. PMID 35663461. [2] Umney, O., et al. (2025). "ClusterNet: Classifying Single-Molecule Localization Microscopy Datasets with Graph-Based Deep Learning of Supracluster Structure." Small Science 5(12): e202500255. PMID 41395524.