Spatially resolved DNA sequencing can reveal how tumour evolution and intra-tumour heterogeneity relate to tissue architecture, but existing methods have been limited by technical complexity and cost. The authors developed Adaptive Resolution Multiscale Spatial DNA sequencing (ARMS DNAseq) to deliver spatially resolved copy number alteration (CNA) profiles from archival tissue sections using ultra-low coverage whole genome sequencing. The approach is intended to be high-throughput and adaptable to user-defined spatial resolution.
ARMS DNAseq derives CNA information from ultra-low coverage whole genome sequencing applied to small, spatially defined regions (tiles) from archival tissue sections. In the reported application the team profiled more than 766 regions (tiles) sampled from 3 patients. The tiled area covered a total of over 300 mm². Per-tile sequencing produced between 1.2 and 2.6 million mapped reads. Tile sizes reported ranged from 0.1 to 0.99 mm², implying sub-millimetre resolution of the copy number maps.
The method is presented as an adaptable multiscale workflow that converts previously low-throughput region-by-region profiling into a scalable spatial CNA assay. The abstract does not provide detailed wet-lab protocols, sequencing platform identifiers, or the precise computational pipeline parameters in the text provided here.
Applying ARMS DNAseq, the investigators reconstructed CNA profiles across the tiled sections and used these profiles to delineate tumour evolution in a spatial context. The spatial maps revealed more tumour subclones than were apparent or fully represented by bulk multi-region whole genome sequencing, indicating that higher-resolution spatial sampling can uncover subclonal complexity that bulk approaches may obscure.
The source reports that ARMS DNAseq enabled direct spatial copy number profiling from archival sections and that these spatial CNA maps were used to interpret patterns of tumour evolution across the sampled tissue area.
The study reports associations between identified tumour subclones and tissue morphology. In addition, the authors demonstrate that deep learning-derived image representations can be used to predict subclone identity, linking histological image features to underlying copy number-defined subclonal structure. The abstract does not include model architectures, training set sizes, performance statistics, or validation metrics for the image-based predictions; those details were not reported in the source abstract.
To demonstrate multi-omic integration, ARMS DNAseq-derived spatial CNA maps were aligned with spatial transcriptomic data. This integration revealed subclone-specific immune cell co-occurrence patterns and transcriptional programmes that cross subclone boundaries, indicating that transcriptional states and immune contexture can be spatially linked to copy number-defined subclonal architecture.
The source text does not supply specific gene signatures, immune cell types, statistical measures of co-occurrence, or exact methods used for alignment and integration in the abstract provided.
The authors present ARMS DNAseq as a scalable workflow suitable for archival tissue sections. Key claimed advantages include the ability to profile hundreds of spatially resolved regions cost-effectively using ultra-low coverage whole genome sequencing, flexibility in user-defined tile size and resolution, and direct generation of spatial CNA maps that support evolutionary and multi-omic analyses. This converts prior low-throughput, region-by-region approaches into a higher-throughput spatial CNA assay according to the abstract.
The abstract highlights the method and key findings but does not report a number of technical and methodological details in the provided text. Missing details include explicit laboratory protocols, sequencing platform and chemistry, computational steps and code availability, performance benchmarks and validation metrics against orthogonal assays, per-sample clinical annotations, and quantitative results for the deep learning and spatial transcriptomic integrations. These specifics were not reported in the source abstract.
The authors declared no competing interests. Funder information reported in the source includes Prostate Cancer UK and Cancer Research UK among others. The preprint citation indicates a posting date of August 27, 2026. The source notes the preprint is available under a CC-BY-NC 4.0 International license.