🧬 Lecture 6 – Protein Structure Bioinformatics & Classification

(Fun, detailed, and beginner-friendly walkthrough)

This lecture moves from sequence bioinformatics into the world of structure bioinformatics — and that’s where things get very interesting. Instead of comparing strings of amino acids, we compare 3D shapes of proteins.

I’ll go through everything step-by-step and explain the logic behind it clearly.

🧩 1. Structure Alignment vs Structure Superposition

You may hear these terms used interchangeably — but they are not the same.

🔹 Structure Alignment

• Used for different proteins
• They may:
  • Have different sequences
  • Be homologs from different organisms
  • Be mutants
  • Have different residue numbering
• Goal: 👉 Find the best overlap in 3D space

Importantly:

• Not all parts of proteins necessarily overlap
• You must identify which regions should be compared
• Alignment is based on atomic coordinates, not sequence similarity

🔹 Structure Superposition

• Used for identical proteins
• Example: same protein in two conformations
• Goal: Compare structural differences

So:

• Alignment = comparing different proteins
• Superposition = comparing the same protein in different states

🧪 2. How Alignment Works in Practice (PyMOL Example)

In PyMOL, two commands are commonly used:

• align
• super

Example from the lecture: Two proteins (CPZ and CC8) both had:

• 4-stranded β-sheet
• 2 α-helices

After alignment:

• RMSD = 2.2 Å
• That is considered quite good

📏 What is RMSD?

RMSD = Root Mean Square Deviation It measures the average distance between aligned atoms.

• Low RMSD → structures are very similar
• High RMSD → poor alignment

Important:

PyMOL did NOT align sequences — it aligned atomic coordinates

This is structural comparison, not sequence comparison.

🧬 3. Sequence → Structure → Function (Direction Matters!)

The lecture emphasizes something crucial:

🧭 Direction of Determination

1. DNA sequence → determines protein sequence
2. Protein sequence → determines protein structure

BUT NOT THE OTHER WAY AROUND.

Why?

• Genetic code is redundant
• Many DNA sequences → same protein sequence
• Many protein sequences → same structure
• Fewer possible structures than sequences

🔽 Variability decreases downward:

DNA variability > Protein sequence variability > Protein structure variability

This is extremely important for understanding why structure classification works.

🧠 4. Structure vs Function — Not Always Simple

Two important evolutionary insights:

🔹 Similar structure, different function

Rare but possible

🔹 Same function, different structures

Example: serine proteases Different folds, same catalytic activity Likely convergent evolution

This tells us: Structure and function are often linked — but not guaranteed.

🗂 5. Why Classify Protein Structures?

Because:

• There are fewer possible folds than sequences
• We want to organize structural space
• Helps predict function
• Helps understand evolution

Two major databases attempt this:

🏛 6. CATH Database

CATH stands for:

• Class
• Architecture
• Topology
• Homologous superfamily

🔹 Level 1: Class

Four main classes:

• Mainly α
• Mainly β
• α/β
• Few secondary structures (often unstructured proteins)

🔹 Level 2: Architecture

Describes overall arrangement of secondary structures.

Examples in α/β class:

• Super roll
• β barrel
• Two-layer sandwich
• Three-layer sandwich
• αβα three-layer
• ββα three-layer

Architecture = overall 3D arrangement (Not sequence order yet)

🔹 Level 3: Topology

Topology = 👉 The path the polypeptide chain takes through the structure

It depends on:

• Order of secondary structures
• Number of elements

Two proteins can:

• Have same architecture
• But different topology

Because the order in sequence differs

Topology ≠ spatial arrangement only Topology = connectivity pattern

🔹 Level 4: Homologous Superfamily

Groups proteins with:

• Structural similarity
• Evolutionary relationship

🏛 7. SCOP Database

SCOP = Structural Classification of Proteins

More complex hierarchy than CATH.

Levels include:

• Class
• Fold
• Superfamily
• Family

Example: In class α and β:

• 147 different folds

Difference between α/β and α+β is not always obvious

🧩 8. Domain Classification — The Complicated Part

Important: Both CATH and SCOP classify domains, not entire proteins

What is a domain?

A structural and functional unit within a protein.

Example: Pyruvate phosphate dikinase

• SCOP identified 3 domains
• CATH identified 6 domains

Even more confusing:

• Some domains overlap
• Some domains consist of non-contiguous sequence regions

This makes domain definition:

• Difficult
• Sometimes manual
• Not fully reproducible

CATH:

• Automated + manual inspection

SCOP:

• Manual classification

That introduces operator bias.

🔎 9. Finding Similar Structures – The DALI Server

Suppose you:

• Solved a new structure
• Or built a homology model

How do you know if similar structures exist?

👉 Use the DALI server

Process:

1. Upload PDB file
2. DALI compares your structure to all known structures
3. Returns similar hits

Example: Uploaded small copper-binding protein (COPSET)

Returned hits:

• Copper transporting proteins
• Mercury transporting proteins
• Heavy metal binding proteins

Conclusion: Proteins with similar structure often share similar function

🧠 Big Conceptual Takeaways

🧬 1. Structure is more conserved than sequence

You can lose sequence similarity and still retain fold.

🧱 2. Alignment is geometric, not sequence-based

Atomic coordinates are compared.

🏛 3. CATH and SCOP organize structural space differently

• CATH: hierarchical & semi-automated
• SCOP: more manual & detailed

🧩 4. Domain definition is not trivial

It is partly subjective and complex.

🔍 5. DALI is your structural BLAST

It finds structural neighbors.

📌 Conceptual Flow of the Lecture

1. Structural alignment basics
2. RMSD interpretation
3. Sequence → structure → function direction
4. Evolutionary implications
5. Structural classification systems
6. Domain complications
7. Structural similarity search