# Which secondary structure prediction algorithms don't query the PDB/DSSP for known homologous sequences?

There are several secondary structure prediction algorithms. I am interested in the algorithms which predict secondary structure without accessing the PDB to cheat by using the DSSP or STRIDE entry for known homologous structures.

Since DSSP and STRIDE are not performing predictions, rather they analyse the PDB structure to determine whether each residue is a helix, strand or coil. This in turn means that, the 'so called' secondary structure prediction algorithm, also, is not performing predictions (except, of course, in the cases where no homologous structure/sequence is found in the PDB).

Which secondary structure prediction algorithms don't cheat by using DSSP or STRIDE for homologous structures?

Here is a list of secondary structure prediction algorithms, which are accessible via a consensus secondary structure prediction server called @NPS (currently down: https://npsa-prabi.ibcp.fr/NPSA/npsa_seccons.html)

DPM
DSC
GOR1
GOR3
HNN
MPSA
PHD
PREDA
SOPM

• Have you tried googling for ab initio secondary structure prediction? – Devon Ryan Jul 1 '18 at 17:57
• @DevonRyan It is a good idea, and doing so does provide some useful information. However, ab initio modelling is not what I am interested in. Use of knowledge is fine, only it should be generalised, not a specific query to the PDB and DSSP for a specific sequence - in such case, no prediction is performed, only cheating i.e. an ~100% accurate answer. – Aalawlx Jul 1 '18 at 18:45
• If using knowledge is fine the querying PDB and such when possible is in no way cheating. – Devon Ryan Jul 1 '18 at 19:15
• Yes it is cheating. Not performing a prediction = cheating. – Aalawlx Jul 2 '18 at 8:37
• You have a warped definition of cheating if it includes using the most reliable information. Since you implicitly don't want the most likely results, what is your end goal with all of this? – Devon Ryan Jul 2 '18 at 9:08