A novel variable neighborhood search approach for cell clustering for spatial transcriptomics.

Area of Science:

• Computational Biology
• Bioinformatics
• Data Science

Background:

• Cell clustering is crucial for understanding biological systems.
• Existing methods often struggle to integrate diverse data types like gene expression and spatial information.
• Integer Linear Programming has been used for clustering but presents computational challenges.

Purpose of the Study:

• To introduce a novel cell clustering approach using the Variable Neighborhood Search (VNS) metaheuristic.
• To develop a method that clusters cells based on both gene expression and spatial coordinates.
• To extend clustering capabilities beyond cell-type identification to spatial domain analysis.

Main Methods:

• Formulated cell clustering as an Integer Linear Programming minimization problem.
• Developed and applied a novel Variable Neighborhood Search (VNS) based model.
• Integrated gene expression matrices and spatial coordinates for clustering.

Main Results:

• The VNS-based approach effectively navigated complex cell clustering challenges.
• The method demonstrated superior performance compared to existing cell clustering techniques.
• Successfully achieved both cell-type and spatial domain clustering.

Conclusions:

• The proposed Variable Neighborhood Search (VNS) method offers an advanced approach to cell clustering.
• This adaptable algorithm can orchestrate clusters using gene expression and spatial data.
• The methodology has potential applications in biomedical research and spatial data analysis.