CNTools: A computational toolbox for cellular neighborhood analysis from multiplexed images.

Area of Science:

• Computational Biology
• Single-cell Analysis
• Spatial Transcriptomics

Background:

• Cellular neighborhoods are increasingly recognized as critical factors in biological processes, including the development of diseases such as cancer and diabetes.
• Accurate and efficient identification of these neighborhoods from high-resolution cellular data is essential for biological discovery.

Purpose of the Study:

• To develop a comprehensive computational toolbox, CNTools, for end-to-end cellular neighborhood analysis.
• To introduce novel methods for cellular neighborhood identification and post-identification smoothing.

Main Methods:

• Development of CNTools, a toolbox integrating state-of-the-art identification methods and smoothing techniques for annotated cell images.
• Implementation of the novel Cellular Neighbor Embedding (CNE) method and Naive Smoothing technique.
• Application and evaluation of CNTools on three real-world CODEX datasets.

Main Results:

• The Cellular Neighbor Embedding (CNE) method combined with Naive Smoothing demonstrated superior performance compared to existing approaches.
• Quantitative and qualitative evaluations on CODEX datasets confirmed the effectiveness of CNE with Naive Smoothing.
• The analysis revealed more robust biological insights into cellular neighborhoods.

Conclusions:

• CNTools provides a powerful platform for cellular neighborhood analysis, enhancing the understanding of spatial single-cell data.
• The CNE method with Naive Smoothing is a highly effective approach for identifying and analyzing cellular neighborhoods.
• Guidance is provided for selecting appropriate identification and smoothing techniques based on specific dataset characteristics.