Skip to content

ChopChopMF: Workflows & Tips

Task-based recipes. Each one assumes you already have a structure open in ChimeraX and the ChopChopMF panel visible. Cross-references to the concept pages (Confidence Metrics, AlphaMissense) explain why a step matters, not just how to click it.

1. First look at a fresh AlphaFold model — pLDDT coloring

Beginner

  1. Open/fetch your structure with Fetch PDB → AlphaFold2 (by UniProt ID, or via ChimeraX's built-in Fetch/Search/Predict panel that ChopChopMF opens alongside its own).
  2. Select the model in the dropdown and run pLDDT Coloring.
  3. Read the color, not just the number: dark orange = low confidence, blue = high confidence. See Confidence Metrics for what the pLDDT bands (>90 / 70–90 / 50–70 / <50) actually mean biologically before you interpret orange regions as "wrong."
Advanced: per-residue table

Run AlphaSync Residue Analysis on the same model to get a full per-residue table: pLDDT, SASA, RSA, disorder prediction, secondary structure, and contacts — all in one place, instead of eyeballing colors. Useful when you need to justify a residue selection quantitatively (e.g. "keep only residues with pLDDT > 70 and RSA < 0.25") rather than by color alone.

2. Is this variant likely pathogenic? — AlphaMissense mapping

Beginner (human protein)

  1. Fetch PDB → AlphaMissense DB, enter the UniProt ID (e.g. P38398 for BRCA1).
  2. The structure opens and is colored by AlphaMissense score automatically.
  3. Read What is AlphaMissense before treating a "likely pathogenic" color as a clinical conclusion — it isn't one.
Advanced (non-human protein)

AlphaMissense scores only exist for human proteins. For a mouse/zebrafish/other ortholog, use Missense Alignment instead of the direct fetch:

  1. Load your non-human structure.
  2. Select model/chain, enter the human ortholog's UniProt ID.
  3. Run the alignment. Matched (conserved) residues get the human AlphaMissense score; unmatched residues are shown in yellow — not scored.
Expert: how much do you trust an extrapolated score?

Only conserved residues inherit a score, and single isolated conserved residues surrounded by divergent sequence are a weak basis for extrapolation — a mismatch a few residues away can mean the local structural/chemical environment differs enough that the human pathogenicity score doesn't transfer cleanly. Treat extended, contiguous conserved stretches (e.g. a whole conserved domain) as reasonably reliable, and isolated single-residue matches as suggestive at best. This mirrors the general caution in AlphaMissense about not over-reading any single score in isolation.

3. Trimming a model before downstream analysis

Beginner

Modify Structure → Crop Structure: select model/chain, enter the residue range(s) to keep (e.g. 1-120,150-200), click Hide Deletion Preview first to sanity-check the selection, then execute.

Deletions are terminal

Crop and Delete Chain cannot be undone by ChopChopMF's Undo button or ChimeraX's normal undo. If there's any chance you'll want the full model again, run Duplicate Structure first (with an XYZ offset so the copy doesn't overlap) and crop the duplicate.

Advanced: why crop before a homology/structure search

Before a Foldseek search (next workflow), crop away flexible termini, linkers, and other clearly disordered regions (check pLDDT / disorder prediction from AlphaSync first — very low, "spaghetti" pLDDT regions are usually intrinsically disordered rather than mispredicted folded domains). Leaving them in dilutes the structural similarity signal Foldseek is matching on and can push real homologs further down the ranked list.

4. Finding structural homologs — Foldseek

Beginner

Foldseek gives you a GUI for structural homolog searches — "closest structural homologs" to your model, without leaving ChimeraX:

  1. Select the target database: PDB (the default — searches experimental structures) or AlphaFold DB / afdb50 (searches predicted structures).
  2. Select the model and chain to search with (e.g. 1:A); click ↻ Refresh model list if your structure isn't listed yet.
  3. Click ChopChop Foldseek — ChimeraX runs the structural search and reports the closest structural homologs.

Three caveats that materially change your results

The first two are stated directly by the tool's own documentation:

  • "Foldseek searches will be more precise and efficient if you crop away all unnecessary residues within your structure, so the focus is on your Domain of Interest." — see workflow 3 for how to crop.
  • "Disordered parts of the protein you might also want to delete, since IDPs have no fixed structure and are rather an ensemble of possible structures — Foldseek can't map those on a certain structure well." Check pLDDT first (workflow 1) to identify which regions are likely disordered before deciding what to crop away.
  • Don't crop down to a single helix or sheet. A lone secondary-structure element is a generic building block found in countless unrelated folds — searching with one produces noisy, largely meaningless hits. Crop a complete domain (a self-contained fold, typically dozens to a few hundred residues) — small enough to exclude flexible material, large enough to still be a distinctive fold. See Structure vs. Sequence Homology for why this distinction (rigid domain vs. isolated SSE vs. disordered region) matters so much for function inference from a structural hit.
Advanced: PDB vs. AlphaFold DB — which to search

Searching PDB finds experimentally solved homologs — useful when you want a structure with known ligands, crystallographic waters, or functional annotations already attached. Searching AlphaFold DB (afdb50) finds predicted homologs across a vastly larger swath of sequence space (including proteins with no solved experimental structure at all) — useful for finding distant relatives in poorly characterized organisms, at the cost of every hit being itself a prediction (check its pLDDT before trusting it as a structural reference).

5. Reading confidence in a multimer/complex — PAE Contacts

Beginner

  1. Requirement: only one model open in ChimeraX for this tool. Open the AlphaFold-predicted complex .cif and its matching .json.
  2. Analyze Structure → PAE Contacts: pick the two chains you care about, set a distance cutoff (5 Å is a reasonable default; avoid going above ~8 Å or you'll pull in noisy, low-confidence pairs).
  3. Run it — pseudobonds are drawn between residue pairs the model is confident are in contact. This is a structural view of the same information a PAE heatmap gives you numerically — cross-check the two if the interface matters for your conclusions.
Advanced

Follow up with PAE Interaction Residues (after running PAE Contacts) to auto-select and show the contact residues as sticks — useful for figures or for feeding a residue list into mutagenesis planning.

Expert: when to reach for the external PAE Viewer instead

ChopChopMF's PAE Contacts is fast for a quick "does this interface look real" check inside ChimeraX. For a full 2D PAE heatmap across the entire complex (all chain pairs at once, not just two at a time) use the standalone PAE Viewer alongside it — it's better suited to spotting which domains pack confidently against each other in a large multimer, before you drill into specific chain-pair contacts back in ChimeraX.

6. Characterizing a protein-protein interface — PDBePISA + ΔG

Beginner: getting the XML file

PDBePISA analysis needs an XML file you export yourself, from the PDBePISA website (not ChopChopMF):

  1. Open PDBePISA and press the Launch PDBePisa button.
  2. Submit your structure: enter your PDB ID, or upload your .pdb file under Select Coordinate file, and click Analyze.
  3. Interface List: click the Interfaces button to see the list of identified macromolecular contacts.
  4. Detail View: press Details for the specific interface you want to analyze.
  5. Configure Display: scroll to Interfacing residues, set Display level to Residues.
  6. Press the XML button and save the .xml file.

Advanced: mapping interfaces in ChopChopMF

  1. Select the target model — the model the XML's residues actually belong to (e.g. Model 1); click ↻ Refresh model list if it isn't listed.
  2. Click Select PDBePISA XML File and load it — this alone already selects and colors the interface residues in darkorange.
  3. To (re-)apply the full 3-way scoring, click ChopChop PISA Interfaces and select the corresponding _output.defattr file.
  4. Residues are automatically categorized as:
    • Buried — hidden in the interface
    • Hydrogen Bond — specific polar interactions
    • Salt Bridge — electrostatic interactions between charged side chains
Expert: ΔG (solvation energy) coloring

Same XML, but a separate panel focused on thermodynamics rather than contact geometry:

  1. Select the target model, then Load PDBePISA XML File to bring in the data.
  2. You can load more than one XML and switch between them via the Active XML file dropdown.
  3. Toggle Append Mode to accumulate/compare residues across all loaded XMLs at once, rather than one interface at a time.
  4. Leave Neutral band (±ε) checked (default) to keep near-zero-ΔG residues colored light grey as a visual baseline instead of noise.
  5. Use the slider to set a ΔG Cutoff threshold (e.g. 0.50 kcal/mol); with Only show residues ≥ cutoff checked, everything below that is excluded from the view.
  6. Click ChopChop ΔG Coloring to apply the energy-based coloring to the structure.
  7. Click Plot ΔG Values to open a bar chart, scatter plot, and value list for the same data.

Residues with ΔG = 0.0 or zero buried surface area are excluded automatically. This workflow is aimed at identifying interface "hotspots" — good candidates for alanine-scanning-style follow-up experiments.

7. Symmetric assemblies from a density map

Beginner

  1. Measure Center on your map/volume → copy the printed XYZ coordinates from the ChimeraX Log.
  2. Symmetry Copies: paste those coordinates, select your structure model, choose the symmetry group (C2, D3, ...), and generate.

Quick reference: pitfalls

Symptom Likely cause Fix
Crop/Delete removed too much and can't be undone No duplicate made first Always Duplicate Structure before a destructive edit if the original might still be needed
PAE Contacts tool won't run / gives odd results More than one model open Close/hide other models — only one is supported
AlphaMissense fetch shows nothing Protein/organism has no native scores Use Missense Alignment against a human ortholog instead of the direct DB fetch
Foldseek hits look noisy or miss the obvious homolog Flexible/disordered regions still in the model Crop them out first (see workflow 3), then search (workflow 4)
Foldseek finds nothing for a poorly characterized organism Searched PDB only Switch the target database to AlphaFold DB (afdb50)
ΔG plot dominated by tiny values Neutral band toggle off Leave Neutral band (±ε) checked to hide near-zero residues