# protein secondary structure matching/comparison scoring

I want to compare two secondary structures of aligned proteins. I do not know exactly how to do it well.

Example:

EEEEEEEEEEETTTTTTHCTTTEEECTTTEEECTTECTCCCHHHHHHHHHCHCCTHHEEEETTTEEECTTTEEECTTTEEC

EEEHHHHHHHTTTTTTTTCTTHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHEEEECCCCCCCCCCEEEEEEEEEEEEEE


Example of matching:

EEEEEEEEEEETTTTTTHCTTTEEECTTTEEECTTECTCCCHHHHHHHHHCHCCTHHEEEETTTEEECTTTEEECTTTEEC
|||        |||||| |||                    |||||||||     ||||||      |   |||    ||
EEEHHHHHHHTTTTTTTTCTTHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH----HHEEEECCCCCCCCCCEEEEEEEEEE


Hypothetically: 50% aligned - > score 0.5 - > thats scoring is wrong:

EEEEEEEEEEETTTTTTHCTTTEEECTTTEEECTTECTCCCHHHHHHHHHCHCCTHHEEEETTTEEECTTTEEECTTTEEC
||||||||||||||||||||||||||||||||||||||||||
EEEEEEEEEEETTTTTTHCTTTEEECTTTEEECTTECTCCCHEEEEEEEEEEEEEEEHHHHHHHHHHHHHHHHHHHHHHHH


This case will have similar score...

This is a puzzle for me, I don't know how to approach it technically.

• Normally 3D structures of protein are compared and the RSMD is reported. How did you get your SS, with JPred or from PDBs? Comparing SS might result in some biases: if one of the two protein is getting used as a template for the other say or if both are from PDBs but the SS was called different (if so, use DSS on both). Also note that JPred is like 75% accurate at best. Dec 15 '19 at 21:49
• @MatteoFerla I suspect that these are predicted secondary structures. Dec 16 '19 at 10:02
• @James Exactly my worry (I see a helix of a single AA in a coil). The solution is straightforward, but I am not sure how one could add error estimates to said solutions, which seems like a must... Dec 16 '19 at 10:15
• I suppose using data from JPred, PSIPRED, SPIDER2 and doing the 9 pairwise alignments would suffice on a push. Dec 16 '19 at 10:20
• @MatteoFerla I agree. In lieu of a structure, usually, a meta-prediction would suffice. I think MTG should add why the sequences are being aligned. I think different tools should be used depending on if they are checking to see how much secondary structure prediction error is in a sequence compared to a reference sequence, or if they are checking secondary structure as a proxy for structural divergence. Dec 16 '19 at 10:28