Bayesian modeling of spatial molecular profiling data via Gaussian process

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Gene expression patterns (mRNA, proteins) are crucial for cell function.
• Spatial molecular profiling technologies offer new insights into tissue context.
• Identifying genes with spatial patterns requires advanced computational methods.

Purpose of the Study:

• To develop a novel computational method for analyzing spatial transcriptomics data.
• To identify genes exhibiting distinct spatial expression patterns within tissues.
• To enhance the understanding of molecular mechanisms underlying cell functions.

Main Methods:

• A Bayesian hierarchical model was developed.
• The model incorporates a zero-inflated negative binomial distribution for count data.
• Bayesian inference framework enables robust parameter estimation.

Main Results:

• The proposed model demonstrates improved stability and robustness.
• It shows competitive accuracy compared to existing methods.
• Successful application in simulation studies and real-world data.

Conclusions:

• The novel Bayesian model is effective for spatial transcriptomics data analysis.
• It provides a robust approach for identifying spatially patterned genes.
• This method advances the field of spatial biology and bioinformatics.