🧬 Lecture Summary — Building Atomic Models from Electron Density

🌊 From Diffraction Pattern → Electron Density → Atomic Model

After collecting diffraction data and performing Fourier synthesis, the experimental result is:

➡️ A 3D electron density map of the unit cell

This map is the true experimental representation of where electrons (and therefore atoms) are located. But it is not yet a structure.

👉 The scientist must now interpret the density and build an atomic model inside it.

🎨 Different Ways to Represent Protein Structures (Models)

A “model” can be shown in many formats — each emphasizes different biological or physical insights.

🌀 Cartoon (Ribbon) Representation

• Shows secondary structure elements
• Helps visualize helices, sheets, topology
• Often combined with ball-and-stick residues for important sites

🌍 Surface Representation

• Displays domain organization
• Shows pockets, interfaces, ligand accessibility
• Useful for functional interpretation

🧱 Ball-and-Stick in Electron Density

• Shows how atoms fit into the experimental density
• Used during model validation

📚 Ensemble Models (Typical for NMR)

• Many overlaid models
• Shows flexibility and dynamic regions
• Regions with spread = more mobile

🥚 Atomic Displacement Ellipsoids (Anisotropic B-factors)

• Each atom drawn as an ellipsoid
• Shape indicates direction and magnitude of motion
• Small ellipsoids → rigid region
• Large ellipsoids → flexible region

🧱 How Do We Actually Build the Model?

🦴 Step 1 — Build a Skeleton (Historical Method)

• Skeleton = simplified representation of continuous density
• Helps trace C-alpha backbone path
• First used in the 1970s

Goal: ➡️ Identify how the polypeptide chain winds through the density

🪄 Step 2 — Pattern Builder (Baton Method)

A baton tool is manually placed in the density:

• Connect Cα → next Cα → next Cα
• Creates a C-alpha trace
• Quickly reveals:
  • α-helices
  • β-strands
  • turns

Once backbone is traced → side chains are added.

🧩 Recognizing Secondary Structure from Density

When the backbone trace is known:

You can identify:

• 🌀 α-helix
• 📄 β-sheet (parallel / antiparallel)
• 🔁 turns

These structural motifs have distinct geometric patterns.

⚗️ Side Chain Chemistry Matters

The final model must make:

• Physical sense
• Chemical sense
• Biological sense

Interactions to consider:

• Hydrogen bonds
• Hydrophobic packing
• Polar interactions

Incorrect chemistry = incorrect structure.

🔎 Recognizing Directionality in Density

Peptide bonds produce characteristic density features:

• Carbonyl “bumps”
• Planar peptide geometry
• Side chains emerge from Cα

These features allow you to determine:

➡️ N-terminus → C-terminus direction of the chain.

🧬 Identifying Specific Amino Acids in Density

Some residues are especially diagnostic:

⭐ Very characteristic residues

• Glycine → no side chain
• Proline → cyclic backbone link
• Methionine → sulfur density
• Aromatics → large rings (Phe, Tyr, Trp)

Scientists often:

• Use primary sequence knowledge
• Search for distinctive density patterns Example: two adjacent tryptophans.

⚡ Using Heavy Atoms and Anomalous Scatterers

Helpful tricks:

• Mercury binds cysteine → reveals cysteine positions
• Selenium-methionine labeling → shows methionine sites
• Sulfur anomalous maps → locate cysteines/met

These provide anchor points for building the model.

🔮 Secondary Structure Prediction Helps!

Before model building, researchers often:

• Predict helices/sheets computationally
• Compare prediction with observed Cα trace
• Helps assign sequence register correctly

📊 Resolution — The Key Quality Indicator

Resolution determines how detailed the density is.

🟥 ~4 Å (Low resolution)

• Mostly featureless
• Only fold / chain path visible

🟧 ~3 Å

• Some side chains visible

🟩 ~2 Å

• Hydrogen bonding visible
• Waters / ions visible
• Good model quality

🟦 ~1 Å (Atomic resolution)

• Individual atoms visible
• “Full chemistry” interpretation possible

🔁 Rotamer Libraries — Fixing Side Chains

Side chains adopt preferred conformations.

Rotamer libraries:

• Statistical database of allowed conformations
• Weighted by observed frequency
• Helps quickly fit density

Alternative:

• Manually drag atoms into density
• Then refine geometry computationally.

🧮 Difference Fourier Maps — Extremely Powerful Tools

These maps show what is missing or wrong in the model.

Fo − Fc map

• Positive density → something missing
• Negative density → something incorrectly modeled

Typical color convention:

• Green → add atoms
• Red → remove atoms

Applications:

• Place water molecules
• Identify mutations
• Locate metal binding sites
• Detect conformational changes

❌ Common Model Building Errors (Very Important for Exams)

From worst → mildest:

🚨 Severe

• Completely wrong fold
• Backbone traced in wrong density
• Secondary structures connected incorrectly

⚠️ Moderate

• Wrong chain direction
• Out-of-register sequence placement

🙂 Minor

• Wrong peptide plane flip
• Wrong side chain rotamer

Even PDB structures can contain errors → always be critical.

🤖 Automated Model Building (Modern Practice)

Examples:

• RESOLVE (Phenix)
• ARP/wARP

👍 Advantages

• Fast
• Objective
• Can build 50–90% of model

👎 Limitations

• May fail in difficult regions
• Hard to define molecular boundaries
• Trouble with ligands / nucleic acids / modifications

Manual finishing is still essential.

🧠 Big Picture — What This Lecture Wants You to Understand

Protein structure determination is not:

❌ “Software gives structure automatically”

It is:

✅ A scientific interpretation process

You must:

• Understand density features
• Know chemistry of residues
• Use Fourier maps intelligently
• Validate stereochemistry
• Be skeptical of models