CLaSSiNet: A Computational Framework for High-Resolution Classification and Spatial Mapping of Heterogeneous...

Area of Science:

• Cell Biology
• Biophysics
• Computational Biology

Background:

• Super-resolution fluorescence imaging, like single-molecule localization microscopy (SMLM), faces challenges in analyzing sparse and discontinuous biological molecular networks.
• Existing methods struggle to resolve higher-order organization due to obscured connectivity, periodicity, and symmetry in SMLM data.

Purpose of the Study:

• To develop a novel computational framework, Classifier of Super-resolution Structural Networks (CLaSSiNet), to overcome sparsity and heterogeneity constraints in SMLM data.
• To enable sensitive capture and automated mapping of network organizational signatures at high resolution.

Main Methods:

• Integration of connectivity, 1D periodicity, and 2D regularity classifiers using newly developed algorithms.
• Development of CLaSSiNet to segment and map networks, resolving four distinct organizational states (1D periodic, 2D polygonal, disordered, non-network).

Main Results:

• CLaSSiNet successfully mapped organizational heterogeneity in the actin-spectrin membrane-associated periodic skeleton (MPS) with unprecedented resolution (~256 nm).
• Revealed ordered MPS networks at cell edges/junctions and non-network states in the cell body.
• Uncovered mechanical coupling between actin stress fibers and spectrin lattices, demonstrating bidirectional coordination.

Conclusions:

• CLaSSiNet provides a robust platform for analyzing SMLM data, regardless of labeling chemistry.
• The study reveals novel organizational principles and mechanical coupling in MPS networks, with cell-specific variations.
• Establishes a principled computational framework for dissecting nanoscale design rules of complex biological and bioinspired architectures.