In our lab, we are working on different projects involving 10X Genomics single-cell and spatial transcriptomics data from 5-30 patient samples each. Our aim is to decode the tumour heterogeneity in cervical and gastric cancers. Furthermore, we work with in-house image-based data constituting the ERC Consolidator project recently awarded to our lab, which needs GPU for image processing and downstream analysis. Specific downstream computational methods are built upon the machine and deep learning and require GPU power for computation. Although currently our lab has access to the GPU partition on Dardel, its hardware constitutes AMD GPU. Most transcriptomics packages for downstream analysis are reliant on NVIDIA-CUDA-torch systems and thus are incompatible with AMD. Examples of such packages include cell2location by Bayraktar lab (https://github.com/BayraktarLab/cell2location) and HRCHY-CytoCommunity by huBioinfo (https://hrchy-cytocommunity.readthedocs.io/en/latest/Installation.html). We require these packages for spatial deconvolution and hierarchical clustering, which will enhance the resolution of the tumour stroma found in our data. Thereby, we apply for 250 GPU-hours per month on the Arrhenius partition.