Optimal Gene Panel Selection for Targeted Spatial Transcriptomics Experiments

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics offers insights into tissue microenvironments and cell communication.
• Current technologies face limitations in spatial resolution or gene coverage, often relying on pre-selected gene panels.
• Optimal gene panel design is crucial for maximizing the utility of spatial transcriptomics data.

Purpose of the Study:

• To introduce ReconST, a novel computational method for automated optimal gene panel design in spatial transcriptomics.
• To leverage single-cell RNA sequencing (scRNA-seq) data for informed gene selection.
• To improve the accuracy and effectiveness of spatial transcriptomics profiling.

Main Methods:

• ReconST utilizes a gated autoencoder model to identify optimal gene subsets from scRNA-seq data.
• The method leverages existing scRNA-seq datasets to inform gene panel design.
• Performance was benchmarked using a high-coverage mouse brain MERFISH dataset.

Main Results:

• ReconST demonstrated superior performance compared to existing methods in reconstruction accuracy.
• The method effectively preserved spatial patterns in transcriptomic data.
• ReconST successfully identified optimal gene panels for spatial transcriptomics profiling.

Conclusions:

• ReconST provides a valuable and broadly applicable tool for designing optimal gene panels for spatial transcriptomics.
• This approach significantly enhances the utility of spatial transcriptomics in diverse biomedical research areas.
• Automated gene panel design can overcome limitations of current spatial transcriptomics technologies.