Prediction
of Secondary and Tertiary Protein Structure
Because the sequence of amino acids in a polypeptide ultimately determines its secondary and tertiary structures, investigators have examined the structures found in proteins and tried to relate them to the individual amino acids.
Secondary Structure
- Table 6.6 lists the relative
probabilities that a particular amino acid will form an 
-helix, 
-sheet, and a "turn" in
proteins. Note that the top group of amino acids favors -helices, the middle group favors -sheets, and
the last group favors turns.
The Chou-Fasman rules for predicting secondary structure of a region of a polypeptide sequence are the following:
1. Any segment of 6 residues or more with an
2. Any segment of 5 residues or more, with
3. Tetrapeptides with an
Figure 6.28 shows the amino acid sequence of bovine pancreatic trypsin inhibitor (BPTI). More importantly, Figure 6.28 compares the secondary structures observed in the native protein and predicted by the Chou-Fasman rules. Note that exceptionally good agreement is found between the predicted and observed structures.
Tertiary Structure
- Attempts to predict tertiary structure of proteins have not
been as successful as those for predicting secondary structure.
Folding of sequences depends critically on specific side chain
interactions, often far removed from one another in the amino
acid sequence. Attempts to predict tertiary structure include
efforts to recognize overall patterns in tertiary folding combined
with the prediction of secondary structure. These efforts have
led to the successful prediction of an /-barrel structure
for tryptophan synthase,
which is in excellent agreement with the structure determined
by x-ray diffraction.
Free Energy Minimization - Another approach to predicting tertiary structure uses computer simulation to minimize free energy in spontaneous folding. The computer rotates random-coil chains through a very large number of small permutations in its conformation, calculates the energy of each, and seeks an energy minimum. A simple random search, however, runs into the Levinthal problem (see here). Several approaches have been taken towards simplifying the search. One of these, called Local Independently Nucleated Units of Structure (LINUS) breaks the problem down by first identifying local regions likely to have a particular secondary structure and then folds them together to minimize the energy.
1. NNPredict - Protein Secondary Structure Prediction
2. Molecular Analysis of Protein Structure
4. Multiple Alignments, Sequence Motifs and Structure Inference