Task index#
Find a guide, runnable notebook, or API for a specific operation.
Fundamentals, input, and output#
Task |
Guide or tutorial |
API and notes |
|---|---|---|
Build a default |
Database API; I/O API. Most callers receive packed types and ordering through a build context. |
|
Build a pose from PDB |
|
|
Select a residue range from PDB |
|
|
Build a pose from Biotite or mmCIF |
Preferred general input path; see the I/O API. |
|
Build from OpenFold, RoseTTAFold2, or AtomWorks tensors |
Use Atom37 or map another named layout to canonical tensors; see the model input tutorial. |
|
Preserve chain gaps and disconnected regions |
Keep internal gaps disconnected rather than silently turning them into chemical termini. |
|
Batch heterogeneous poses |
Use |
|
Export Biotite, one PDB, or multiple models |
PDB is not a lossless replacement for CIF plus authoritative ligand chemistry. |
|
Build a stable chemistry subset or inspect packed block-type metadata |
Keep the source |
Kinematics and minimization#
Task |
Guide or tutorial |
API and notes |
|---|---|---|
Build an automatic multi-chain, gap-aware forest |
|
|
Construct explicit or per-residue-root forests |
|
|
Select named torsions and jumps |
Configure a |
|
Run Cartesian or kinematic minimization |
The coordinate models differ. Compare only with matched masks, weights, budgets, and stopping checks. |
|
Run Cartesian, kinematic, or batched FastRelax |
|
|
Couple chains with an explicit jump or construct a dandelion forest |
Validate edge coverage and contiguous ordinary-jump indices before coordinate operations. |
|
Perturb rigid-body jump DOFs or assign named torsions |
These are coordinate transformations, not docking, minimization, or idealization protocols. |
Scoring and constraints#
Task |
Guide or tutorial |
API and notes |
|---|---|---|
Build default, empty, or focused score functions |
||
Score a pose or backpropagate through coordinates |
Render a module for the current pose layout and call it with coordinates. |
|
Analyze weighted or unweighted block pairs |
Directed accounting can require both matrix orientations for an unordered pair. |
|
Map a protein interface and test selected alanine substitutions |
Compose author-label masks, both block-pair orientations, and matched local-repacking tasks. |
|
Reweight an interface differentiably |
Apply an explicit analytical weight tensor before summing and backpropagating. |
|
Add distance, coordinate, or torsion constraints |
See the constraint API. |
|
Construct a focused score function or derive modified score parameters |
Change weights to alter term contributions; derive a new immutable database only when changing the underlying parameter model. |
Packing, design, and preparation#
Task |
Guide or tutorial |
API and notes |
|---|---|---|
Construct samplers and repack a fixed sequence |
|
|
Optimize polar-hydrogen chis or build supported side chains |
Use normal preparation or explicitly configure the relevant sampler. |
|
Add extra χ sampling |
TMol χ indices are zero-based: |
|
Run regional design or a small mutation-score experiment |
Compose explicit task masks. |
|
Subclass |
Candidate enumeration is deterministic for a fixed task; annealing is a separate stochastic assignment search. |
|
Prepare and inject ligand parameters |
Start from authoritative CIF/MOL2 chemistry. |
|
Score controlled ligand-pose decoys and locally refine diagnostic states |
Reuse one ligand-aware context, batch matched rigid-body decoys, and compare scores with ligand displacement. |
|
Score or pack DNA/RNA |
The packer samples glycosidic and hydroxyl-proton chi; sugar pucker stays fixed. |