Tuesday, November 25, 2008

Secondary Structure Prediction

It allows us to find where secondary structural elements (α helices, β sheets, loops) are located.

Secondary Structure Predictors

  1. Chou-Fasman
  2. COMBINE
  3. GORIII
  4. PRISM
  5. PHD0
  6. PHD3
  7. PSIPred

PSIPred uses most recent algorithm that can predict secondary structure of about 80% accuracy.

Protein Threading or Protein Fold Recognition

Although proteins are of large no, tertiary structural motifs are limited to which most protein belongs. It is speculated that about 1000 distinct protein folding patterns may be present in total. Surprisingly, a few dozen folding patterns account for about half of all known protein structures. This helps to use previously solved structures as starting point.

To identify the fold the sequence is compared with all 500+ folds in library of known protein structure.

If pair-wise alignment shows less than or nearly 30% identity then it is ideal to be used for protein fold recognitions.

When successful, structure from fold recognition may be about 3-6Å RMSD (root mean square deviation) from actual structure.


Problems:

  • Only 70% chance there that top 10 prediction contain correct fold.
  • To reduce the number of predictions one requires more information like functional information (as function depicts structure), motifs, and position of exposed residues.
  • Still there is 30% chance that none of top prediction correct.
  • Quality result heavily depends on amount of information from other methods and human expertise.


Previous


Wednesday, November 19, 2008

Protein Structure Prediction

Why prediction of protein structure important? You may ask what is there in protein structure??

That is because structure determines function of the protein. Take the example of enzyme dehydrogenases. It has an NAD-binding site called Rossaman fold ( Dinucleotide binding fold). This fold is made up a pair of βαβαβ subunit. Thus if a protein contain a βαβαβ subunit than it acts as a binding site for a nucleotide.

Structure is better conserved than sequences in protein during the course of evolution e.g. take the example of Cytochrome C of eukaryotes and C-cytochromes of prokaryotes in different species(which change with evolution in prokaryotes) where all perform general functions i.e. electron carrier. But different species exibit low degree of similarity of sequence to each other and to that of eukaryotes. They also differ in polypeptide loops on surface. But X-Ray structure are similar particularly chain folds and side chain packing to interior.

We require structural knowledge for rational drug design, protein engineering, for detail study of protein-biomolecule interactions.


Experimental methods to find structure of protein

X-Ray Crystallography: In order to know the structure using it we first have to crystallize the protein. We must have 20mg of material to start with. The results produced by it are accurate.

NMR: Can't be used for structures with more than 120 residues. Protein must be soluble and requires about 30mg/ml. It can locate flexible/rigid regions.

Other methods are Cryo-EM(electron microscopy),
CD(Circular dichroism)

X-Ray Crystallography, NMR, Cryo-EM gives 3D information of proteins but CD only gives one dimensional structure of protein i.e. secondary structure only

Why do we want predicted methods ??

Computer based prediction are much easier to handle with. E.g. one is free to make errors with out compensating much as it is inexpensive at least in future. Moreover we don't have always sufficient material for experimental methods. Some proteins even don't crystallize. So we turn to predicted method if experimental methods fail.

Computation Method of structure prediction

  1. Secondary structure prediction
  2. Protein Threading or Fold Family Recognition
  3. Ab-initio structure prediction
  4. Homology Modeling