Model SanctumModel Sanctum
In-app help — Model Sanctum 1.0.0, quick reference, deploys with the platform. Product docs (MkDocs) — full guide at docs.int.bayer.com, deploys independently via CI.

What is Model Sanctum?

A GPU-accelerated platform for running computational molecular models — protein structure prediction, binder design, molecular docking, immunogenicity scoring, and more. No GPU setup, no Docker commands, no cluster management. Run on cloud GPUs or route jobs to on-premise HPC clusters via the built-in Slurm bridge.

Why "Sanctum"?

A sanctum is a sacred, protected space. Model Sanctum is exactly that — a dedicated space where computational models are preserved, maintained, and made accessible. Instead of models scattered across laptops, threads, and one-off scripts, they live in one curated environment: containerized, versioned, GPU-ready, and available to everyone. The name reflects our philosophy — treat models as first-class artifacts that deserve proper infrastructure, not afterthoughts.

Structure Prediction

ColabFold, Boltz, IntelliFold

Protein Design

RFdiffusion, BindCraft

Molecular Docking

DiffDock

ADMET & Properties

ADMET, GenMol

Immunogenicity

BayMune (DigiMAb)

Sequence Analysis

MSA Search, iPSAE

Quick Start

1

Log in

Go to modelsanctum.sbx.powerml.int.bayer.com and sign in with Bayer SSO.
2

Pick a model

Go to Tools and choose the model you want to run (e.g. Boltz for structure prediction).
3

Submit a job

Fill in the form (paste a protein sequence, upload a file, etc.) and click Submit.
4

Monitor & download

Watch real-time logs on the job page. When complete, view 3D structures in the built-in viewer or download output files.

Key Concepts

JobA single model run. Goes through: submitted → queued → running → completed/failed.
ProjectGroups jobs together, like a folder. Share projects with team members.
Model / ToolA containerized computational tool (ColabFold, Boltz, etc.).
PipelineChain multiple models together (e.g. MSA → Fold → Dock).
Compute TargetWhere to run: Cloud GPU (EKS) or HPC Slurm (on-prem).