Skip to content

Cookbook

Recipes for getting things done with molforge. Each recipe answers a specific task-oriented question — "I want to do X, what do I write?" — and shows a complete, runnable example. The User guide covers concepts; this section covers concrete workflows.

If you want to...

Task Recipe
See the whole Identity layer in one script (design → reproduce) End to end
Predict a structure from a sequence Fold a sequence
Predict a protein + ligand or multi-chain complex Multi-component cofolding
Dock a small molecule against a folded receptor Fold then dock
Get a raw PDB ready for MD simulation Prepare for MD
Run a short MD simulation and analyse it MD and RMSD
Design sequences for a backbone, then validate by re-folding Design then refold
Trace what produced an output across a multi-step workflow Inspect provenance
Check a folded or docked structure for quality problems Validate structures
Skip recomputing expensive engine calls you've already run Caching results
Rank a series of analogs by binding affinity (MM/GBSA) Rank binders with MM/GBSA
Rank analogs by rigorous relative affinity (FEP, via alchemlyb) Rank binders with FEP
Compute an absolute binding free energy (ABFE, via alchemlyb) Absolute binding free energy with FEP
Ingest, clean, dedup, and filter a set of small molecules Work with small molecules
Fetch or search structures (RCSB/AlphaFold) and compounds (ChEMBL) Fetch and search databases

If you're choosing between options...

Decision Comparison
Which folding engine should I use? Folding engines
Which docking engine should I use? Docking engines
Which generative engine for what task? Generative engines

How these recipes work

Every recipe is structurally complete — real imports, real method signatures, real arguments — and will run as written if you have the dependencies for the engine it uses. Most recipes need optional extras:

  • Folding via ESMFold needs pip install "molforge[ml]", plus torch and a few GB of weights.
  • Docking via Vina needs the vina Python package and Open Babel.
  • MD via OpenMM needs pip install "molforge[md,prep]" and a working OpenMM install.

Each recipe states its requirements at the top so you know what you're in for before you copy the code.

For shorter, more conceptual introductions, see the walkthroughs. For exhaustive worked examples that combine multiple engines, see the examples notebooks.