Principles of Protein Structural Organization - H. Schultz 1982

Prediction of secondary structure from amino acid sequence
Probabilistic methods
Propensity of three residues to incorporate into secondary structure

An attempt to predict the overall folding of a chain of a related protein group using triplets. Kabat and Wu used triplets of adjacent residues in the sequence to predict not only the Secondary Structure [360], but also the folding of the entire chain [361, 362]. To predict chain folding, the angles (∅, ψ) of the central residues of all triplets in Proteins with known three-dimensional structures forming the baseline set were recorded as a "(∅, ψ) table", which included 20 ∙ 20 ∙ 20 = 8000 positions, one for each triplet type.

The method was applied to predict the overall chain folding of a group of related proteins (Cytochromes c or IMMUNOGLOBULINS, Sec. 9.1). In such a group, many Amino acid sequences are characterized by a similar chain fold, i.e., identical angles (∅, ψ). Therefore, for a given residue at position i, triplets (i − 1, i, i + 1) can be compiled across all homologous proteins, and their angles ∅ψ* can then be obtained using the (∅, ψ) table. The averages of these angles (φ, ψ) are chosen as the predicted values for residue i. The predicted chain folding is obtained by repeating this Procedure for all residues.

* Let us consider the following hypothetical case: four related proteins have the following sequences at triplet positions (i - 1, i, i + 1): Ala-Gly-Thr, Val-Gly-Pro, Leu-Ala-Ser, and Leu-Ala-Thr. For each type of triplet, the corresponding table position includes a set of ∅, ψ angles, namely a set of occurrence frequencies for the one under consideration