How to Align Protein Structures in PyMOL: align, super and cealign Explained

The three alignment commands at a glance

align — sequence-guided alignment

align first performs a sequence alignment between the mobile and reference structures, maps the sequence-aligned residues onto each other, then minimizes the RMSD by rotating and translating the mobile structure. After the initial fit, it iteratively removes the worst-fitting residue pairs and recalculates — the reported RMSD reflects the well-fitting core rather than the whole protein.

# Basic alignment: mobile structure aligns to reference
align mobile, reference

# Output example:
 Executive: RMSD =    0.742 (197 to 197 atoms)
 Executive: RMSD =    0.598 (183 to 183 atoms) after outlier rejection

# Align specific chains
align mobile and chain A, reference and chain A

# Align and store the result object (for examining the alignment)
align mobile, reference, object=alignment_result

# Align without moving the mobile structure (get RMSD only)
align mobile, reference, cycles=0, transform=0

PyMOL prints two RMSD values: the initial alignment RMSD (all matched pairs) and the refined RMSD after outlier rejection. The refined value is typically reported in publications. The number of atoms used appears in parentheses — a significantly lower atom count in the refined alignment suggests many poorly-matching regions were rejected, which can indicate the alignment is unreliable.

super — structure-guided alignment

super skips the sequence alignment step entirely and works directly from structural coordinates. It identifies structurally equivalent residue pairs by comparing local geometry — backbone angles, secondary structure patterns — rather than sequence similarity. This makes it reliable across a much wider range of sequence identities.

# Structure-based alignment — preferred for distant homologs
super mobile, reference

# Output example:
 Executive: RMSD =    1.243 (201 to 201 atoms)
 Executive: RMSD =    0.891 (187 to 187 atoms) after outlier rejection

# super on backbone atoms only (faster, cleaner RMSD)
super mobile and backbone, reference and backbone

# super on specific chains
super mobile and chain B, reference and chain A

cealign — fold topology alignment

cealign implements the Combinatorial Extension algorithm, which searches for the longest continuous path of structurally similar fragments between two proteins. It’s the most computationally intensive of the three but is specifically designed for the hardest cases: proteins with similar folds but very different sequences, or structures where the topological arrangement of secondary structure elements differs between the two proteins.

# CE algorithm alignment — for distantly related or topologically similar proteins
cealign reference, mobile

# Note: argument order is reversed from align and super!
# cealign REFERENCE, MOBILE (not mobile, reference)

# CE alignment and store the transformation
cealign reference, mobile, object=cealign_result

When to use each — decision guide

Interpreting RMSD values

RMSD (root mean square deviation) measures the average distance between equivalent atom pairs after alignment. Lower values mean the structures are more similar. What counts as “good” depends entirely on the context — comparing crystal structure to AlphaFold prediction is different from comparing two conformational states of the same protein.

Aligning multiple structures

When comparing several structures simultaneously — a set of crystal structures in different conformations, or five AlphaFold models — align them all to a single reference:

# Load multiple structures
fetch 4XYZ 5ABC 6DEF 7GHI

# Align all to 4XYZ as reference
super 5ABC, 4XYZ
super 6DEF, 4XYZ
super 7GHI, 4XYZ

# Color each structure distinctly to compare
color slate,    4XYZ
color salmon,   5ABC
color palegreen, 6DEF
color lightorange, 7GHI

# Show all as cartoons
as cartoon
zoom

For an automated multi-structure alignment in a script, use a Python loop within PyMOL. This is practical when aligning more than four or five structures:

structures = ["5ABC", "6DEF", "7GHI", "8JKL"]
reference  = "4XYZ"

for s in structures:
    cmd.fetch(s)
    cmd.super(s, reference)

Aligning specific domains or regions

When two proteins share one well-conserved domain but differ in others — kinases with conserved catalytic cores and variable regulatory domains, for example — aligning on the whole protein produces a poor overall superposition that obscures the conserved region. Aligning on the specific domain of interest is almost always more informative.

# Align only on the kinase domain (residues 200-450)
super mobile and resi 200-450, reference and resi 200-450

# Align on specific chain of a complex
super mobile and chain A, reference and chain A

# Align on backbone of secondary structure elements only
super mobile and ss h and backbone, reference and ss h and backbone

# Align on a named selection defined earlier
select core, resi 50-150 and name CA
super mobile and core, reference and core

Calculating RMSD without moving structures

Sometimes you want the RMSD value without actually repositioning any structure — when the structures are already in a meaningful orientation, or when you want to measure deviation between pre-aligned models.

# RMSD of backbone atoms without moving anything
rms_cur mobile and backbone, reference and backbone

# RMSD of Cα atoms specifically
rms_cur mobile and name CA, reference and name CA

# align without transformation (RMSD only, no movement)
align mobile, reference, transform=0