Lecture 5

Question 1

The SCOP database

Answer 1

Structural Classification of Proteins Database developed in Cambridge

• Sorted proteins based on structural classes->folds->Superfamilies->families.

For example: Small proteins->Cysteine-knot cytokines->Cysteine-knot cytokines->Transforming growth factor beta - more of a functional than structural classification

a-helical

B-sheet

a+B - a-helices and B-sheets in different parts of proteins, no B-a-B motifs

a/B - Helices and sheets assembled from B-a-B motifs

a/B-linear - Line through centres of strands of sheet roughly linear

a/B-Barrels - Line through centres of strands of sheet roughly circular

Proteins with little/no secondary structure e.g. soft proteins

Question 2

What does SCOP struggle with?

Answer 2

Domains

Many proteins are multi-domain and SCOP assumes they are single-domain

For example: clotting factors like factor XII:
Fn2-EGF-EGF-Fn1-Kr-SerPr

Factor IX:
Gla-EGF-EGF-SerPr

Difficult to classify evolutionary origins due to domain shuffling.

Question 3

How do we define domains

Answer 3

The Gō plot method

Defines domains using distances of amino acids from centre of a protein and whether they cluster distantly to the centre of a protein.

Conducting Go plot:
1. Calculate the radius of spherical volume of protein
2. Calculate distance from each a-carbon of each amino acid to all the others
3. If distance is greater than spherical radius, score +.

Lines are drawn in a triangle to identify protein domains

Never get completely clean triangles in real proteins

Question 4

What are the disadvantages of Gō plots

Answer 4

• Requires solved structures
• Domain boundaries not always clear
• Gō method now superseded by sequence based algorithms

Question 5

Explain modular evolution of proteins

Answer 5

• Domain boundaries are usually at exon boundaries
• Not all exon boundaries are domain boundaries
• Genome rearrangements are important in evolution of new domain combinations

Question 6

How does Pfam build domains?

Answer 6

• Start with high quality protein structure (X-ray crystallography, good resolution, low angstrom).
• BLAST PDB to find related protein structures.
• Align these - maximise structural homology (adjust alignment so secondary structural element boundaries match).
• Build a statistical profile (Hidden Markov Model - HMM) of ‘seed’ alignment.

Question 7

Gaining a protein from PDB

Answer 7

Input seuqence and Protein sequence databank -> BLAST search

BLAST search -> Filter results (E<threshold) -> Multiple alignment sequence -> Position-specific scoring matrix

Question 8

Hidden Markov modelling

Answer 8

Sequence

Question 9

Showing patterns using sequence logos

Answer 9

Sizes of DNA/amino acids can be used to illustrate their presence across a series of different molecules.

Question 10

What software shows DNA/amino acid patterns using sequence logos

Answer 10

Pfam on a larger scale

Question 11

How does pfam build domains?

Answer 11

Use the HMM to query GenPept – hmmsearch
Align the new hits to the HMM – hmmalign
Rebuild the HMM to include the new hits – hmmbuild
Repeat as desired, or until there are no new hits
“Structure, structure, structure” (Alex Bateman, founder of Pfam)

Question 12

Disadvantages of Pfam?

Answer 12

• Domains defined by a HMM, and HMM only as good as ‘seed’ alignment used to construct it.
• HMM building process is iterative, so errors can be magnified.
• Curation is uneven due to numbers of domains in Pfam
• Viruses under-represented

Question 13

Homstrad

Answer 13

Homologous Structure Alignment Database

Used to collect good seed alignments

Used in construction of globin molecules

Question 14

Not enough structures

Answer 14

• Structure determination is far harder than sequencing
• Illumina and Minion sequencing have made sequencing ultra high throughput
• No equivalent technological leap forward for structural biology

The structural genomics consortium