Coot Refinement

Coot Refinement Best Practices

This skill provides guidance for effective and safe structure refinement in Coot, based on lessons learned from crashes and workflow optimization.

Critical Configuration

Set Immediate Replacement Mode for Scripting

ALWAYS call these two before any refinement operations in scripts:

coot.set_refinement_immediate_replacement(1)
# Makes refinement synchronous — results are committed immediately.

coot.set_imol_refinement_map(imol_map)
# Tells Coot which map to refine against.
# CRITICAL: without this, refine_residues_py() silently fails (returns -2 with no atoms moved).
# Call once per session, or whenever the active map changes.

Why: Enables synchronous operation where refinement directly updates coordinates. Without this:

• Refinement happens asynchronously in background threads
• Need to call coot.set_refinement_immediate_replacement() before using refinement functions (just once is enough) which should remove the risk of threading conflicts and crashes, race conditions between refinement and rendering.

When to use: Any time you're scripting refinement operations (refine_residues_py, refine_zone, etc.)

Safe Refinement Workflow

1. Enable immediate replacement

coot.set_refinement_immediate_replacement(1)

2. Set the refinement map

coot.set_imol_refinement_map(imol_map)

3. Refine residues (preferred method)

Pass a list of residue specs: [["A", 42, ""], ["A", 43, ""], ...]

residue_specs = [["A", resno, ""] for resno in range(start_resno, end_resno + 1)] result = coot.refine_residues_py(imol, residue_specs)

4. so-called "sphere refine" is often useful because it allows movement/improvement

of residues that are close in space but distant in sequence.

central_residue_spec = ['A', 12, ''] neigbs = coot.residues_near_residue_py(imol, central_residue_spec) residue_spec = neigbs residue_specs.append(central_residue_spec) result = coot.refine_residues_py(imol, residue_specs)

Hard-Mode-Soft-Mode-Hard-Mode

Some problems are of the kind where the geometry restraints "get in the way" of the atoms moving to the correct minimum. In such cases, you can try letting the refinement "go soft" - ie.

o Mediocre Refinement o increase the map weight (weight_matrix) by a factor of 10 or so. o Refine the same residues again o Restore the previous map weight o Refine the same residues again

This can sometimes be helpful to remove bad geometry or bad fit - atom clashes in particular.

Additionally, or perhaps at the same time, you can try to reduce LJ epsilon using coot.set_refinement_lennard_jones_epsilon() by a factor of 100 or so. And then restore it and re-refine the same residues.

Zone Refinement

# Enable immediate replacement
coot.set_refinement_immediate_replacement(1)

# Refine zone
coot.refine_zone(
    imol,
    chain_id,
    start_resno,
    end_resno,
    ""  # alt conf
)

Validation and Correlation

Check Residue Quality

# Get density correlation for specific residue
correlation = coot.density_correlation_analysis_scm(imol, chain_id, resno, ins_code)
# Returns dict with 'all-atom' and 'side-chain' correlations

# Get worst residues
worst = coot.get_n_residues_with_worst_density_fit(imol, n_residues)
# Returns list of [chain_id, resno, inscode, correlation]

Correlation Targets

Good fit:

• All-atom correlation: > 0.8
• Side-chain correlation: > 0.8

Poor fit (needs attention):

• All-atom correlation: < 0.5
• Side-chain correlation: < 0.5

Common Crash Scenarios to Avoid

Threading/Rendering Conflicts

Problem: Rapid-fire refinement operations while graphics are rendering can cause memory corruption in GTK rendering pipeline.

Symptoms:

nanov2_guard_corruption_detected
gdk_gl_texture_new_from_builder
gtk_gl_area_snapshot

Solution: Use set_refinement_immediate_replacement(1) for synchronous operation