GANANA (www.ganana.eu) is a Europe-India partnership for Scientific High-Performance Computing aiming to leverage the power of HPC to progress on global challenges within life sciences, geophysical hazards, and climate and weather. Within the life science, we work on accelerating the GROMACS molecular dynamics software and developing a framework for computational AI-driven immunogenic peptide design. Targeted immunotherapy that uses short peptides to elicit an immune response to eliminate infected or cancerous cells, offers a promising therapeutic approach for immunologic diseases. Here, an immune response is initiated when the Major histocompatibility complex (MHC) and T-cell receptor (TCR) recognize and bind to a peptide fragment from a foreign protein - either originating from an infectious agent or a cancer cell - leading to activation of T-cells that attack the defect cell. However, identifying peptide epitopes and structural properties of the receptors that will lead to an immunogenic response is challenging. The purpose of our project is therefore to develop a large-scale computational workflow involving multiple computational methods such as deep-learning of receptor and peptide sequences, structure prediction, molecular docking, and molecular dynamics simulations and free energy calculations, to identify MHC/peptide/TCR combinations with potential immunogenic properties. Within the GANANA collaboration, we are primarily focusing on developing a molecular dynamics/free energy calculation modules capable of ranking the physical viability of the AI-generated peptide/receptor complex designs and to compute useful immunogenic properties, such as binding affinity, from the simulations, which can then be used for further training of the neural network for sequence prediction. Although identifying receptor/peptide combinations capable of eliciting T-cell immunogenic response is a difficult problem, a successfully built workflow could ultimately serve as a tool for epitope-based vaccine development targeting both infectious diseases and certain types of cancer. Preliminary results showed that the MHC-peptide binding is remarkably stable, even when docking scores are poor; and that peptide-MHC-TCR complexes are stable with varying interaction patterns based on the peptide and alleles. This enabled us to identify the molecular dynamics parameter to be integrated in the molecular dynamics modele of computational workflow. Within the GANANA project, we are developing algorithmic innovations to significantly accelerate and improve molecular dynamics simulations in GROMACS. Our work focuses on reducing overhead in memory and computation through the introduction of fixed-precision coordinate representations, as well as modernizing communication algorithms to better exploit emerging hardware architectures. Continuous access to large-scale computational resources is essential for this effort, as it enables systematic performance analysis, scalability studies, and iterative refinement of these developments through ongoing benchmarking campaigns. In preparation for next-generation HPC systems, including the upcoming Arrhenius supercomputer, we request access to the Dardel-GPUH partition based on NVIDIA GH200 nodes. While using the same Slingshot 11 interconnect as Dardel-GPU, the GH200 architecture introduces tight CPU–GPU coupling, offering both opportunities and challenges for communication-focused algorithm development critical to our project.