I studied that prediction of transmembrane alpha helices is more easy and accurate and also good algorithms are available for their prediction. But when we move towards prediction of transmembrane beta barrels (cylindrical structure composed of anti parallel beta-sheets) then for their prediction only a small number of algorithms are present which also don't give reliable results.

My query is that Why it so? Why the prediction of transmembrane beta barrels is difficult and why developers can't develop reliable algorithms for this purpose?

  • $\begingroup$ This might just be me being cynical, but isn't it predominantly to do with them being a lot more rare than TMH? Less cynically I would also say that the hydrophobic signal for a TMH is usually very strong. Perhaps TM beta barrel is more nuanced when compared to non-TM beta barrels? $\endgroup$ – James Feb 23 '16 at 8:52

I could tell only from my pretty poor experience. I had all PDB entries processed with DSSP and then I've tried to find any trends in amino acid preferences. I did not find any except the pretty obvious absence of proline in double (i.e. with two neighbours) strands. Also, negative charged amino acids (I, L, V) were slightly more frequent in parallel strands. After that, I tried a sort of naive Bayes classifier for the amino acid and it's neighbours in two neighbour strands and +/-2 amino acids up/downstream. For single\double parallel\antiparallel strands, the prediction precision was about 60%.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.