MOTIVATION LETTER My research depends on high-performance computing. The neurocascade simulation engine couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics across three calibrated receptor systems, each requiring Bayesian MCMC calibration with 62 passing tests. The TOPOLOGIX pipeline runs ESM-2 protein-language-model delta-embeddings against 553 mutations on the Platinum benchmark, generating AUROC scores of 0.804 plus or minus 0.025. The ergofluids project applies Koopman-operator methods with Mori-Zwanzig memory kernels to model drug transport through non-Newtonian tumor tissue, a computation that failed its first real-data gate and demands more sophisticated HPC approaches to succeed. These three projects, all active in 2025-2026, share a common bottleneck: they need cluster-level parallelization that my current Nextflow and SLURM workflows only partially exploit. The EMBL Bio-IT HPC Cluster Training addresses this bottleneck directly. I have run ODE simulations with RK45 integrators, deployed NEURON and Brian2 for spiking neural network models, and managed job arrays on institutional clusters. What I lack is systematic knowledge of EMBL-specific cluster architecture, parallel I/O optimization, and advanced resource management techniques that would let me scale neurocascade from three receptor systems to ten, or run TOPOLOGIX across the full SKEMPI 2.0 benchmark of 3,085 mutations instead of the 553 I currently cover. The training's emphasis on practical job submission and parallel computing matches my learning style: I need hands-on exercises tied to real research problems, not abstract lectures. My independent researcher status makes this training particularly valuable. Without a university HPC support team, I must self-optimize every computational step. The EMBL training would give me the cluster-specific skills that institutional researchers get from their local administrators. I will apply every technique learned directly to the ergofluids project, which needs to pass its second real-data gate after the first gate failed its pre-registered criterion. Faster, more efficient parallel computation may be the difference between a null result and a meaningful finding. Nigeria has no EMBL-node HPC facility. As a Nigerian researcher training in Germany, I can bridge that gap by learning EMBL's methods and documenting them for African collaborators who face even steeper computational barriers. The training is a force multiplier for my existing work and a foundation for future cluster-dependent projects. SHORT ESSAY: RELEVANCE OF HPC TO MY RESEARCH Three active projects demonstrate why HPC skills are foundational to my work. First, the CCT model of reward-memory encoding prevention in addiction. This tripartite ODE framework with 14 free parameters required Bayesian MCMC calibration using PyMC's DEMetropolisZ sampler. The posterior super-additivity of 13 to 22 percentage points across model versions emerged only after weeks of cluster-calibrated sampling. Without HPC, the model would have converged on local optima or required unacceptable simplifications. I need to extend this model to include serotonin and noradrenaline axes, which will double the parameter count and require four times the compute. Second, TOPOLOGIX predicts drug-resistance mutations from protein-language-model embeddings. The current pipeline covers 553 mutations on the Platinum benchmark. The structure-limited baseline mCSM-lig covers only 18 percent of mutations. TOPOLOGIX covers 100 percent but at a computational cost that scales with sequence length and embedding dimension. To reach the full SKEMPI 2.0 benchmark, I need parallelized embedding generation and distributed Random Forest training across cluster nodes. Third, ergofluids applies Dynamic Mode Decomposition with memory kernels to drug transport. The first real-data gate failed its primary pre-registered criterion. The next gate requires finer spatial discretization and longer simulation times, both of which demand cluster-level parallelization that my current single-node workflows cannot provide. The EMBL training will teach me cluster-specific optimization techniques that apply directly to these three projects. I will learn how to profile parallel I/O bottlenecks, manage memory across distributed nodes, and submit efficient job arrays for parameter sweeps. These skills are immediately transferable to neurocascade, TOPOLOGIX, and ergofluids. SHORT ESSAY: LEARNING GOALS AND EXPECTED OUTCOMES I have three specific learning goals for the EMBL HPC Cluster Training. First, I want to master EMBL-specific job submission and resource management. My current SLURM workflows are functional but inefficient. I submit jobs with conservative resource requests because I do not know how to profile memory and CPU usage accurately. The training's practical exercises on cluster architecture and job scheduling will let me write submission scripts that use 80 percent of allocated resources instead of 40 percent, cutting wall-clock time for neurocascade calibrations by half. Second, I need to learn parallel I/O optimization for large embedding datasets. TOPOLOGIX generates ESM-2 embeddings that are 1,280-dimensional per residue. For a typical protein of 400 residues, that is 512,000 floating-point numbers per sequence. Writing and reading these embeddings inefficiently creates I/O bottlenecks that idle compute nodes. The training covers parallel file systems and I/O strategies that I can implement immediately in my DuckDB-based ingest pipelines. Third, I want to understand distributed memory programming for the ergofluids Koopman-operator computations. The Mori-Zwanzig memory kernel requires solving integral equations that do not parallelize trivially. I need to learn MPI or equivalent approaches to distribute the computation across cluster nodes while minimizing communication overhead. The expected outcome is measurable: within three months of completing the training, I will have reduced the wall-clock time for a full neurocascade calibration run by 40 percent, expanded TOPOLOGIX coverage to 1,000 mutations, and passed the second real-data gate for ergofluids. I will document my optimized workflows and share them with other independent researchers through my personal site zyco.org. CHECKLIST - [ ] Complete EMBL Bio-IT HPC Cluster Training registration form - [ ] Prepare CV highlighting HPC experience with NEURON, Brian2, Nextflow, SLURM - [ ] Write one-page summary of current HPC-dependent projects (neurocascade, TOPOLOGIX, ergofluids) - [ ] Confirm training dates do not conflict with M.Sc. Digital Health coursework at HPI/Potsdam - [ ] Verify eligibility for external researchers (non-EMBL staff) - [ ] Prepare list of specific software dependencies needed for training exercises EDITOR NOTES - Eligibility risk: The training page states it is open to EMBL staff and external researchers, but the registration process may prioritize EMBL affiliates. Eniola should confirm external researcher access before investing time in detailed preparation. - Fact verification: The SKEMPI 2.0 benchmark size of 3,085 mutations needs confirmation from the current database version. The profile states TOPOLOGIX covers 553 mutations on Platinum; the SKEMPI 2.0 number should be verified against the latest release. - Personal detail gap: The profile does not specify whether Eniola has prior experience with EMBL-specific software or cluster environments. If he has used EMBL tools or collaborated with EMBL researchers, that information should be added to the application.