Abstract
Recent advancements in spatial transcriptomics have been largely triggered by two high-resolution technologies: Visium-HD and Xenium in-situ. While sequencing-based Visium HD features a refined bin size of 2 µm and transcriptome wide coverage, Xenium in-situ is a targeted imaging-based detection technology with sub-micron resolution. Herein we use a publicly available lung dataset which contains Visium-HD and Xenium-5K data generated on identical tissue slides to make a bona-fide technical comparison aligned with thorough pathological annotations. Whilst Visium-HD offers a broader gene coverage for detection and likely detects more tumor subclones, Xenium-5K achieves comparable results when robust clustering algorithms are applied. Importantly, from the pathological point of view, the single-cell segmentation accuracy is essential when analyzing irregularly shaped cells, where Xenium may be in favor. At the opposite side, although Xenium-5K based cell segmentation to delineate immune cells, normal lung, and vasculature at cell resolution is decent, it relies on fluorescent signals for transcript detection, which is challenging in heavily pigmented tissues such as melanoma or dust-laden alveolar macrophages, an application scenario for which Visium HD may stand out. From this perspective, pathological derived factors are the prior consideration for selecting an appropriate ST approach under difference research settings including cancer.