Imported from pemsley/coot (mcp/docs/skills/NCS/SKILL.md). Install upstream with npx skills add pemsley/coot --skill NCS. Copyright stays with the author.

Using NCS/Symmetry-Related Molecules as Reference

Overview

When building or rebuilding fragments, use the sequence and model from NCS-related (non-crystallographically related) or molecular symmetry-related molecules as a reference. If one copy is better built, use it to guide building of the other copies.

Why This Helps

Multiple copies of the same molecule in the asymmetric unit should have identical sequences
One copy may have better density or be more complete than others
The better-built copy provides the correct sequence and approximate coordinates

Workflow: Compare Sequences First

Before building, compare sequences between related chains:

def compare_chain_sequences(imol, chain1, chain2):
    """Compare sequences of two chains, return differences."""
    diffs = []
    for resno in range(1, 200):  # Adjust range as needed
        name1 = coot.residue_name_py(imol, chain1, resno, "")
        name2 = coot.residue_name_py(imol, chain2, resno, "")
        
        if name1 and name2 and name1 != name2:
            diffs.append((resno, name1, name2))
        elif name1 and not name2:
            diffs.append((resno, name1, "MISSING"))
        elif name2 and not name1:
            diffs.append((resno, "MISSING", name2))
    
    return diffs

# Compare chains A and B
diffs = compare_chain_sequences(0, "A", "B")
if diffs:
    print("Sequence differences:")
    for resno, a, b in diffs:
        print(f"  {resno}: A={a}, B={b}")
else:
    print("Sequences are identical")

Using Better Chain as Template

If chain B is more complete, use it to guide chain A:

# Chain B has residues 94-96 built, chain A doesn't
# Get sequence from chain B
target_seq = ""
for resno in range(94, 97):
    name = coot.residue_name_py(0, "B", resno, "")
    if name:
        # Convert 3-letter to 1-letter code
        one_letter = {'GLY': 'G', 'ALA': 'A', 'VAL': 'V', 'LEU': 'L', 
                      'ILE': 'I', 'PRO': 'P', 'PHE': 'F', 'TYR': 'Y',
                      'TRP': 'W', 'SER': 'S', 'THR': 'T', 'CYS': 'C',
                      'MET': 'M', 'ASN': 'N', 'GLN': 'Q', 'ASP': 'D',
                      'GLU': 'E', 'LYS': 'K', 'ARG': 'R', 'HIS': 'H'}
        target_seq += one_letter.get(name, 'X')

print(f"Target sequence from B: {target_seq}")

# Build residues in chain A, then mutate to correct sequence
# ... add_terminal_residue calls ...

# Mutate to match chain B
coot.mutate_and_autofit_residue_range(0, "A", 94, 96, target_seq)

Avoiding Common Mistakes

Don't assume "auto" gives correct residue type
- add_terminal_residue(imol, chain, resno, "auto", 1) builds ALA by default
- Always mutate to the correct sequence afterward
Check for subtle differences
- GLY vs GLN both start with 'G' - compare 3-letter codes, not single letters
- Use residue_name_py() not single-letter conversions for comparisons
Verify after mutation
- Check density correlation after mutating
- Use auto_fit_best_rotamer() to optimize side chain placement

Complete Example: Build Fragment Using NCS Reference

def build_fragment_from_ncs(imol, target_chain, ref_chain, start_resno, end_resno):
    """Build missing residues in target_chain using ref_chain as template."""
    
    checkpoint = coot.make_backup_checkpoint(imol, 
        f"before building {target_chain}/{start_resno}-{end_resno} from {ref_chain}")
    
    # Get reference sequence
    ref_seq = ""
    for resno in range(start_resno, end_resno + 1):
        name = coot.residue_name_py(imol, ref_chain, resno, "")
        if name:
            one_letter = {'GLY': 'G', 'ALA': 'A', 'VAL': 'V', 'LEU': 'L', 
                          'ILE': 'I', 'PRO': 'P', 'PHE': 'F', 'TYR': 'Y',
                          'TRP': 'W', 'SER': 'S', 'THR': 'T', 'CYS': 'C',
                          'MET': 'M', 'ASN': 'N', 'GLN': 'Q', 'ASP': 'D',
                          'GLU': 'E', 'LYS': 'K', 'ARG': 'R', 'HIS': 'H'}
            ref_seq += one_letter.get(name, 'X')
    
    print(f"Reference sequence from {ref_chain}: {ref_seq}")
    
    # Find last existing residue in target chain
    last_resno = start_resno - 1
    while coot.residue_name_py(imol, target_chain, last_resno, ""):
        pass
    last_resno -= 1
    
    # Extend chain
    for i in range(end_resno - last_resno):
        current = last_resno + i
        result = coot.add_terminal_residue(imol, target_chain, current, "auto", 1)
        if result != 1:
            print(f"Failed to add residue after {current}")
            break
        
        # Refine as we go
        specs = [[target_chain, r, ""] for r in range(max(1, current-1), current+2)]
        coot.refine_residues_py(imol, specs)
        coot.accept_moving_atoms_py()
    
    # Mutate to correct sequence
    coot.mutate_and_autofit_residue_range(imol, target_chain, 
                                          start_resno, end_resno, ref_seq)
    
    # Final refinement
    specs = [[target_chain, r, ""] for r in range(start_resno, end_resno + 1)]
    coot.refine_residues_py(imol, specs)
    coot.accept_moving_atoms_py()
    
    # Verify
    print(f"\nBuilt residues in {target_chain}:")
    for resno in range(start_resno, end_resno + 1):
        name = coot.residue_name_py(imol, target_chain, resno, "")
        print(f"  {resno}: {name}")

# Usage:
build_fragment_from_ncs(0, "A", "B", 94, 96)

Using NCS Matrices to Copy Fragments

For more substantial rebuilding, you can copy entire fragments from one NCS copy to another using the NCS transformation matrix.

Step 1: Enable NCS Ghosts and Get the Matrix

# Create NCS ghosts (calculates the transformation matrices)
coot.make_ncs_ghosts_maybe(imol)

# Get NCS information
ncs_chains = coot.ncs_chain_ids_py(imol)  # e.g., [['A', 'B']]
ghosts = coot.ncs_ghosts_py(imol)

# Extract the rotation matrix and translation vector
# ghosts[0] = [description, from_chain, to_chain, [rot_matrix, trans_vector], bool]
rot_trans = ghosts[0][3]
rot = rot_trans[0]    # 9 elements: m11,m12,m13,m21,m22,m23,m31,m32,m33
trans = rot_trans[1]  # 3 elements: x,y,z

print(f"NCS transforms {ghosts[0][1]} onto {ghosts[0][2]}")

Step 2: Copy Fragment from Reference Chain

# Delete the problem region in target chain
for resno in range(1, 11):
    coot.delete_residue(imol, "A", resno, "")

# Create a fragment molecule from the reference chain
frag_mol = coot.new_molecule_by_atom_selection(imol, "//B/1-10")

Step 3: Apply NCS Transformation

# Transform the fragment to the target position
coot.transform_molecule_by(frag_mol,
    rot[0], rot[1], rot[2],   # row 1 of rotation matrix
    rot[3], rot[4], rot[5],   # row 2
    rot[6], rot[7], rot[8],   # row 3
    trans[0], trans[1], trans[2])  # translation

Step 4: Change Chain ID and Replace Fragment

# Change chain ID from B to A
coot.change_chain_id(frag_mol, "B", "A", 0, 0, 0)

# Paste into the target molecule
coot.replace_fragment(imol, frag_mol, "//A/1-10")

Step 5: Refine with Flanking Residues

# Include flanking residues for better annealing at the junction
specs = [["A", r, ""] for r in range(1, 14)]  # 1-10 plus 11-13
coot.refine_residues_py(imol, specs)
coot.accept_moving_atoms_py()

Complete Function: NCS Fragment Copy

def copy_fragment_by_ncs(imol, from_chain, to_chain, start_resno, end_resno, flank=3):
    """Copy a fragment from one NCS chain to another using NCS matrix.
    
    Parameters
    ----------
    imol : int
        Molecule index
    from_chain : str
        Source chain ID (the better-built one)
    to_chain : str
        Target chain ID (the one to rebuild)
    start_resno : int
        First residue number to copy
    end_resno : int
        Last residue number to copy
    flank : int
        Number of flanking residues to include in refinement
    """
    
    checkpoint = coot.make_backup_checkpoint(imol, 
        f"before NCS copy {from_chain}/{start_resno}-{end_resno} to {to_chain}")
    
    # Ensure NCS ghosts are calculated
    coot.make_ncs_ghosts_maybe(imol)
    
    # Get NCS matrix
    ghosts = coot.ncs_ghosts_py(imol)
    if not ghosts or ghosts == False:
        print("No NCS ghosts found!")
        return False
    
    # Find the right ghost (from_chain -> to_chain)
    rot_trans = None
    for ghost in ghosts:
        if ghost[1] == from_chain and ghost[2] == to_chain:
            rot_trans = ghost[3]
            break
        elif ghost[1] == to_chain and ghost[2] == from_chain:
            # Need inverse - for now, skip this case
            print(f"Found inverse NCS ({to_chain} -> {from_chain}), may need inverse_rtop_py")
            rot_trans = ghost[3]
            break
    
    if not rot_trans:
        print(f"No NCS relationship found between {from_chain} and {to_chain}")
        return False
    
    rot = rot_trans[0]
    trans = rot_trans[1]
    
    # Delete target residues
    for resno in range(start_resno, end_resno + 1):
        coot.delete_residue(imol, to_chain, resno, "")
    
    # Create fragment from source chain
    atom_sel = f"//{from_chain}/{start_resno}-{end_resno}"
    frag_mol = coot.new_molecule_by_atom_selection(imol, atom_sel)
    
    if frag_mol < 0:
        print(f"Failed to create fragment from {atom_sel}")
        return False
    
    # Transform fragment
    coot.transform_molecule_by(frag_mol,
        rot[0], rot[1], rot[2],
        rot[3], rot[4], rot[5],
        rot[6], rot[7], rot[8],
        trans[0], trans[1], trans[2])
    
    # Change chain ID
    coot.change_chain_id(frag_mol, from_chain, to_chain, 0, 0, 0)
    
    # Replace fragment in target molecule
    target_sel = f"//{to_chain}/{start_resno}-{end_resno}"
    result = coot.replace_fragment(imol, frag_mol, target_sel)
    
    if result != 1:
        print("replace_fragment failed")
        coot.restore_to_backup_checkpoint(imol, checkpoint)
        return False
    
    # Refine with flanking residues
    refine_start = max(1, start_resno - flank)
    refine_end = end_resno + flank
    specs = [[to_chain, r, ""] for r in range(refine_start, refine_end + 1)
             if coot.residue_name_py(imol, to_chain, r, "")]
    
    coot.set_go_to_atom_chain_residue_atom_name(to_chain, start_resno, "CA")
    coot.refine_residues_py(imol, specs)
    coot.accept_moving_atoms_py()
    
    # Clean up fragment molecule
    coot.close_molecule(frag_mol)
    
    print(f"Successfully copied {from_chain}/{start_resno}-{end_resno} to {to_chain}")
    return True

# Usage:
copy_fragment_by_ncs(0, "B", "A", 1, 10, flank=3)

When to Use NCS Fragment Copy

Large missing regions - faster than building residue-by-residue
Poorly built regions - replace with better NCS copy
After molecular replacement - one copy may fit better initially
Symmetric assemblies - enforce consistency across copies

Key Takeaway

When multiple copies of the same molecule exist, use the better-built copy as a reference for sequence and model building. This ensures:

Correct residue types (not default ALA)
Consistent sequences across all copies
Proper disulfide bonds and other features

For larger rebuilds, use the NCS matrix to transform and paste entire fragments - this is faster and more accurate than building from scratch.

coot-NCS-reference-guidance

About