Advancing Spatial Omics Analysis with AINA, the LabCobot, Powered by NVIDIA’s VISTA-2D AI Foundation Model
Veröffentlicht von Eva Hernschier, SupraTix GmbH (2 Monate, 1 Woche her aktualisiert)
AINA, the LabCobot, is bringing new precision to genomics research by leveraging NVIDIA’s VISTA-2D AI Foundation Model to support the analysis of spatial omics data. Genomics researchers use advanced sequencing techniques to understand biological systems, with spatial omics offering insights that go beyond cellular-level data by incorporating spatial context to reveal where each cell resides within tissue. As researchers move towards integrating multiple omics data at the tissue level, spatial omics is transforming how we interpret cellular interactions in development and disease.
AINA, the LabCobot, is bringing new precision to genomics research by leveraging NVIDIA’s VISTA-2D AI Foundation Model to support the analysis of spatial omics data. Genomics researchers use advanced sequencing techniques to understand biological systems, with spatial omics offering insights that go beyond cellular-level data by incorporating spatial context to reveal where each cell resides within tissue. As researchers move towards integrating multiple omics data at the tissue level, spatial omics is transforming how we interpret cellular interactions in development and disease.
For spatial omics to provide this level of insight, imaging is essential. Techniques like fluorescence tagging allow scientists to observe markers without separating cells from their native tissue, keeping spatial information intact. This method helps identify and understand molecules such as RNA and proteins in their exact tissue locations, providing a richer context for analyzing cellular function and disease progression.
AINA supports this paradigm shift by using NVIDIA GPUs and VISTA-2D to process high-density spatial omics data directly onboard. This integration enables AINA to assist researchers in managing data at an unprecedented scale, processing up to ~150 TB per cm² of tissue. By automating and accelerating cell segmentation—a critical first step in spatial omics analysis—AINA ensures that tags are accurately associated with the correct cells, producing precise and reliable data.
VISTA-2D’s transformer-based network architecture, with ~100 million hyperparameters, allows AINA to handle the diverse cell morphologies and scale required for tens to hundreds of thousands of cells in a single sample. Using Meta’s Segment Anything Model (SAM) pretrained weights, VISTA-2D achieves high-resolution, instance-based segmentation, even across different imaging modalities, such as brightfield, phase-contrast, fluorescence, confocal, and electron microscopy.
With AINA’s robust preprocessing and postprocessing pipelines, researchers can now run advanced cell morphology and gene perturbation tasks. AINA, the LabCobot, opens new doors for spatial omics, enabling researchers to transform their understanding of health and disease, accelerating breakthroughs in drug development and spatial diagnostics, and contributing to the ongoing revolution in Life Sciences Spatial Biology.