Workflows

Use these concise recipes for recurring TMol tasks. The Tutorials are executable, notebook-length demonstrations; these pages focus on reusable steps and link to the corresponding Tutorials for deeper analysis.

Prepare structures

• Structure I/O and visualization
  • Choose an input path
  • Build a PoseStack from mmCIF
  • Inspect deposited and built structures
  • Export
  • Common checks
• Model and structure integrations
  • RoseTTAFold2
  • OpenFold
  • Biotite and AtomArray
  • Differentiable AtomWorks Atom37 coordinates
• GPU batching
  • Time CUDA work correctly
  • Chunk larger workloads

Score, design, and refine

• Scoring and analysis
  • CPU batch throughput
  • Ligand-aware Scoring
  • Block-pair Scores
  • Fragmented-ligand attribution
• Packing
• Protein-interface analysis
  • Define partners from metadata
  • Score the interface
  • Batch explicit mutations
  • Interpret the result
• Minimization and FastRelax
  • Cartesian Minimization
  • Constraints
  • Missing Side Chains and Hydrogens
  • Kinematic Minimization
  • Relax

Specialized chemistry

• Ligand Preparation
  • Entry Points
  • Loading Complexes
  • Reuse Prepared Context
  • Persist Prepared Ligands
  • SMILES to Params CLI
  • Interaction Scores
  • Troubleshooting
  • Public API
• Nucleic acids
  • Repack selected bases

Develop and benchmark

• Benchmarking
  • Running Developer Benchmarks
  • Comparing Revisions
  • Profiling
• Development
  • Local Setup
  • Building Extensions
  • AOT vs JIT Extension Loading
  • Tests
  • Containers
  • CI
  • Releasing
  • Code Style

How the workflows interact

Structure preparation determines chemical types, atom layout, and topology. Those choices define the PoseStack consumed by scoring, packing, and minimization. Packing can change block identities and atom counts, so render a new scorer for its returned stack. Coordinate-only minimization can reuse a scorer built for the same layout. FastRelax alternates packing and minimization and handles that transition for its supported schedule.

Ligand and nucleic-acid workflows reuse the same I/O, scoring, packing, and movement primitives with additional chemistry and samplers. GPU batching is an execution strategy that can wrap many of these workflows; it does not make unrelated scores directly comparable.

Use the task index for a direct operation lookup or the interactive examples for complete walkthroughs.