Application note · Specialised

Application Note: Molecular Dynamics and AI-Assisted Virtual Screening

A campaign design joining CPU preparation, GPU simulation, biomolecular models, scheduler arrays and governed scientific evidence.

Virtual screening is a campaign, not a single benchmark. Preparation, conformer generation, docking, molecular dynamics, feature extraction and AI ranking have different resource shapes. The objective is completed, reviewable compounds per day with enough evidence to reproduce each decision.

Campaign architecture

AxiLattice handles CPU preparation, docking variants, ETL and statistical analysis. AxiForge Max runs four-GPU single-node training or simulation. AxiBastion supports multiple concurrent GPU campaigns and distributed training across twelve 96GB devices. Keep structures and approved results on protected shared storage; stage trajectories to NVMe.

GROMACS execution

Benchmark the exact molecular system because CPU/GPU balance changes with atom count, PME configuration and trajectory output. Allocate CPU cores to keep accelerators fed without stealing capacity from concurrent runs. Write checkpoints to resilient storage at a measured interval and test continuation after interruption.

AI ranking

Use versioned descriptors, sequence/structure inputs and splits that prevent scaffold or temporal leakage. BioNeMo or other domain frameworks can support model development, but validation remains specific to the scientific question. Rank candidates with uncertainty and preserve full-simulation confirmation for high-value decisions.

Scheduler strategy

Use job arrays for independent replicas and compounds; reserve multi-GPU allocations for models that demonstrate scaling. Apply project quotas and fair-share so one campaign cannot crowd out all work. Track licence tokens as schedulable resources where necessary.

Evidence pack

For every selected compound preserve input structures, protonation/parameter choices, seeds, software images, checkpoints, model/version, scores and review outcome. Acceptance tests include energy/trajectory sanity, repeatability within expected stochastic tolerance, throughput and recovery.

Primary technical references

References are provided for software architecture and implementation planning. Validate the versions, licences, support matrix and regulated-use requirements applicable to the final deployment.

Relevant Hyperion platforms

Hyperion X1460

AxiForge Max

The maximum single-node Hyperion platform for AI training, molecular modelling, generative engineering and large accelerated workflows.

One 4U node; 128 AMD EPYC cores; 1.15TB ECC DDR5; four NVIDIA RTX PRO 6000 96GB GPUs providing 384GB aggregate VRAM.

Hyperion X3140

AxiLattice

A shared departmental CPU-compute facility for research pipelines, simulation, genomics and queued multi-user work.

Three 4U nodes; 384 AMD EPYC cores; 3.4TB aggregate ECC DDR5; three professional GPUs; switched 100GbE RDMA, 25GbE storage and Slurm.

Hyperion X3460

AxiBastion

A sustained departmental AI/HPC platform for pharmaceutical modelling, large training campaigns and shared accelerated research.

Three 4U nodes; 384 AMD EPYC cores; 3.4TB ECC DDR5; twelve NVIDIA RTX PRO 6000 96GB GPUs; 100GbE compute and 25GbE storage fabrics.

Configuration and quotation

Validate this workload on Hyperion

Final architecture and price depend on representative code and data, concurrency, storage, networking, site infrastructure, component availability and export compliance.

Request Formal Quotation