This algorithm is for prediction of the most stable fragments of alpha helices and beta strands in 3D structures of proteins.
The algorithm works with PDB file.
Open up PDB file by any text editor and copy all its content to the "PDB" list.
Enter the letter designating the chain from PDB file that you want to study in the cell on the "Chain" list.
For models that have no description of secondary structure elements in lines "HELIX" and "SHEET" enter the n umbers of first and last residues in helices and beta strands on the "for models" list.
Copy all the content of the "colored PDB" list to any text editor and save it as the PDB file.
Open up the resulting PDB file by 3D structure viewer and choose the option "temperature" to see the most stable fragments of alpha helices (in green), somewhat stable fragments of alpha helices (in yellow), the most stable fragments of beta strands (in red), and somewhat stable fragments of beta strands (in orange). Remaining parts of a protein are colored in blue.
On the "Predictions" list one will see the table containing predictions in text format (H, E or C).
There are 2 columns:
1) prediction made by the alpha helical pattern (in favor of alpha helices);
2) prediction made by the beta structural pattern (in favor of beta strands).
On the "mono" list one will see predictions by the original amino acid propensity scale in numbers. On the "mono GR" list one will see the plot with probabilities for alpha helix, beta strand and random coil along the length of a sequence.
On the "pentapeptide" list one will see predictions by the original pentapeptide propensity scale in numbers. On the "penta GR" list one will see the plot with probabilities for alpha helix, beta strand and random coil along the length of a sequence.
Probability scales can be found on the "mono" and "pentapeptide" lists.
Four last lists contain probability scales based on dipeptide analysis of the most stable types of beta strands, alpha helices and random coil regions.
Using three options on the "Predictions" page advanced users may try to increase or reduce the influence of dipeptide scales on the final prediction.