Protein structure modelling aided by X-ray scattering and crystallographic measurements

Protein structures can be solved by either experimental measurements such as X-ray crystallography and nuclear magnetic resonance, or computational methods such as homology modeling, fold recognition, and ab initio structural prediction methods. Either way has its limitations. To make the best combination of them, we explore the strategy of combining the computational algorithms with inputs from simple experiments (in particular Small Angle X-ray Scattering measurements or SAXS) to see how the experimental structural information can improve the protein structure prediction. We used a new fitness score of SAXS profile similarity to evaluate and filter sets of candidate structures generated by ab initio methods or threading, in addition to the use of standard energy scores.; Three separate projects are conducted to fulfill this motivation: (1) Lattice model based ab initio protein structure prediction; (2) Dummy residue model based ab initio protein structure prediction; (3) Fold recognition.; We have tested our algorithms on a large set of small proteins and have demonstrated that SAXS is indeed able to improve the results of protein structure prediction, by eliminating false positive candidate structures without favorable SAXS scores or providing new shape and volume constraints to the generation of candidate structures.; Another project is on the comparative study of motor proteins' conformational change: myosin and F1 ATPase versus kinesin. We did normal modes analysis of the elastic network models based on the crystal structures of those motor proteins. We found that the measured conformational change is dominated by one or two low frequency normal modes in myosin and F1 ATPase but not in kinesins, which suggests that ‘mechanical power stroke’ picture may only apply to the former but not the latter which is more consistent with a Brownian random search mechanism.