FICT

FICT assigns cell types in spatial transcriptomics data from fluorescence in situ hybridization (FISH) by integrating gene expression and neighborhood spatial information through a probabilistic model optimized with an Expectation-Maximization algorithm.

Key Features:

• FISH spatial transcriptomics compatibility: Operates on spatial transcriptomics data obtained via fluorescence in situ hybridization (FISH).
• Integration of expression and spatial data: Combines gene expression profiles with spatial context derived from neighboring cells for cell type assignment.
• Probabilistic function optimization: Optimizes a formalized probabilistic function tailored to spatial transcriptomics complexities.
• Expectation-Maximization algorithm: Leverages an Expectation-Maximization algorithm for parameter estimation and assignment refinement.
• Iterative refinement: Iteratively incorporates neighborhood information to refine cell type assignments across successive cycles.

Scientific Applications:

• Cell type and sub-type identification: Assigns cell types and refines sub-type classification in spatial transcriptomics datasets.
• Benchmarking on simulated and real datasets: Demonstrates performance on both simulated datasets and real-world spatial transcriptomics data.
• Neuronal sub-type discovery: Applied to excitatory and inhibitory neurons to identify novel spatial sub-types and their spatial organization.

Methodology:

FICT formulates a probabilistic model that integrates gene expression with neighboring-cell spatial information and optimizes it via an Expectation-Maximization algorithm with iterative refinement of cell type assignments.