Dear colleagues,
I am very beginner in modeling who was lucky enough to face with hard task. I will be very grateful for the helpful advice.
The task:
I have a novel protein with unknown function. This is N-glycosylated secretory protein 158 aa with 5 disulfide bonds (DB), connectivity pattern of which was determined experimentally by MS. The aim is to reveal function of this novel protein. This protein was purified from snail and it is not surprising that no any significant similarity was found among databases using sequence-derived approach. So, we have no homologous template. One possible approach is to find protein with similar fold, in order to get some hints about function. To do this, we need to predict tertiary structure of our protein.
I-Tasser and Quark was used to create models. I have used CA distance restraints 5.6A for Quark and CA 5.6A, SG 2.02A for I-Tasser.
My questions:
1. Why reastraints from DB do not improove TM score of models? Tm score with restarints even lower then without. Why?
QUARK TM-score of Model 1: 0.3128 ±0.0833
I-Tasser 0.26±0.08 (TM-score)
2. I know that for "correct" prediction of protein length N one need N/8 distance restraints. I have a five DB, that is four time smaller number. TM-score of generated models is close to random fold. Can I hope on building of model with correct fold?
3. Among generated models, those with "correct" (by visual inspection) CYS positions was choosen. How to join CYS to form disulfide bonds?
4. Is it necessary to carry out additional processing of models after bonding CYS?