STGBench: sequencing-level spatial DNA-RNA simulation for multimodal and virtual cell-oriented benchmarking of...

Area of Science:

• Genomics and Transcriptomics
• Computational Biology
• Cancer Research

Background:

• Spatially resolved genomics and transcriptomics are crucial for understanding tumor evolution and therapeutic resistance.
• Benchmarking spatial analyses of copy number variation (CNV), single-nucleotide variants (SNV), and mutation burden is limited by a lack of ground-truth datasets.
• Existing simulators often fail to provide matched DNA-RNA outputs or propagate genomic variations to sequencing signals, hindering end-to-end multi-omics pipeline benchmarking.

Purpose of the Study:

• To introduce STGBench, a novel sequencing-level spatial DNA-RNA simulator.
• To enable end-to-end benchmarking of spatial genomics and transcriptomics analysis pipelines.
• To provide a controllable benchmark generator with explicit ground truth for spatial CNV/SNV and mutation-burden analyses.

Main Methods:

• STGBench synthesizes paired DNA-seq alignments (BAM files) and gene expression matrices on a user-defined 2D tissue grid.
• It incorporates spatial CNV landscapes and SNV/VAF fields, coupling copy number states to expression via a negative binomial model.
• The simulator accounts for spatially correlated technical effects and generates outputs directly consumable by downstream bioinformatics tools.

Main Results:

• Spatial CNV profiles were recovered with high accuracy (Pearson r up to 0.855) using AneuFinder on simulated DNA data.
• CNV-driven expression signatures were reproduced (r up to 0.996) using InferCNV on simulated transcriptomes, supporting clonal organization.
• STGBench accurately simulated diverse spatial SNV patterns, with realistic allelic balance and CNV-associated coverage shifts observed.

Conclusions:

• STGBench provides a robust platform for generating realistic spatial DNA-RNA data for benchmarking.
• The simulator facilitates the evaluation of computational tools for spatial CNV, SNV, and mutation burden analyses.
• STGBench offers explicit ground truth across paired DNA-RNA modalities, advancing the field of spatial multi-omics analysis.