The objective of this project is to strengthen evidence-based clinical decision-making in pediatric health care. This will be achieved by integrating biological (gut microbiome), behavioral, and clinical data to develop and validate prediction models for long-term ketogenic diet treatment outcomes in 100 children with drug-resistant epilepsy. In parallel, the project aims to identify microbial species linked to seizure reduction that can be used for the development of live biotherapeutic products (LBPs) as novel anti-seizure therapies. These species may, in combination with a KD, enhance treatment responsiveness in partial or non-responding patients, or enhance adherence and decrease the everyday burden of a ketogenic lifestyle by reducing the strictness of the KD while maintaining effectiveness. We will shotgun sequence fecal samples from 100 children collected before and after starting KD- Will plan to use Uppmax resources for the following: Metagenomic data will be processed using the StaG metagenomic workflow collaboration (mwc) pipeline (https://stagmwc.readthedocs.io/en/latest) for taxonomic and functional annotations. State-of-the-art machine learning approaches will be applied to build prediction models of KD treatment response based on baseline microbiome composition. Algorithms will include Partial Least Squares Regression, Elastic Net generalized linear models, Neural Networks with dimensionality reduction, Ordinal Random Forests, k-Nearest Neighbors, eXtreme Gradient Boosting, and Support Vector Machines with radial basis function kernels. Models will undergo rigorous cross-validation, hyperparameter tuning, and stepwise feature selection. Model performance will be carefully evaluated using appropriate metrics. Additional clinical and behavioral variables (e.g., age, sex, epilepsy subtype, medication profile, dietary habits) will be incorporated to assess whether combined clinical-biobehavioral models improve predictive accuracy. Microbial species with the highest predictive value will be ranked by importance. In addition, therapy-associated microbiome changes will be identified by comparing baseline and three-month samples and correlating microbial abundance changes with individual seizure reduction. Taken together, the project seeks to enable individualized treatment planning, sustained self-management, and improved quality of care for children with severe epilepsy and their families through high-throughput sequencing and application of ML algorithms which requires high performance computing services. I am uncertain of the exact computing resources this project will require and gratefully accept your advice.