(XFEL) allows it to outrun radiation damage in coherent diffractive imaging since elastic scattering terminates before photoelectron cascades commences. This “diffract-before-destroy” feature of XFEL opened up new opportunities for biological macromolecule imaging and structure studies by breaking the limit to spatial resolution imposed by the maximum dose that is allowed before radiation damage. However, data collection in serial femto-second crystallography (SFX) using XFEL is affected by a bunch of stochastic factors, which pose great challenges to the data analysis in SFX. These stochastic factors include crystal size, shape, random orientation, X-ray photon flux, position and energy spectrum. Monte-Carlo integration proves effective and successful in extracting the structure factors by merging all diffraction patterns given that the data set is sufficiently large to average out all stochastic factors. However, this approach typically requires hundreds of thousands of patterns collected from experiments. This dissertation explores both experimental and algorithmic methods to eliminate or reduce the effect of stochastic factors in data acquisition and analysis. Coherent convergent X-ray beam diffraction (CCB) is discussed for possibilities of obtaining single-shot angular-integrated rocking curves. It is also shown the interference between Bragg disks helps ab-initio phasing. Two-color diffraction scheme is proposed for time-resolved studies and general data collection strategies are discussed based on error metrics. A new auto-indexing algorithm for sparse patterns is developed and demonstrated for both simulated and experimental data. Statistics show that indexing rate is increased by 3 times for I3C data set collected from beam time LJ69 at Linac coherent light source (LCLS). Finally, dynamical inversion from electron diffraction is explored as an alternative approach for structure determination.
This dissertation begins by proposing a new flexible policy iteration (FPI) framework. To improve sample efficiency, FPI can utilize either on-policy or off-policy learning strategy, can learn from either online or offline data, and can even adopt exiting knowledge of an external critic. Approximate convergence to Bellman optimal solutions are guaranteed under mild conditions. Simulation studies validated that FPI was data efficient compared to several established RL methods. Furthermore, a simplified version of FPI was implemented to learn from offline data, and then the learned policy was successfully tested for tuning the control parameters online on a human subject.
Next, the dissertation discusses RL control with information transfer (RL-IT), or knowledge-guided RL (KG-RL), which is motivated to benefit from transferring knowledge acquired from one subject to another. To explore its feasibility, knowledge was extracted from data measurements of able-bodied (AB) subjects, and transferred to guide Q-learning control for an amputee in OpenSim simulations. This result again demonstrated that data and time efficiency were improved using previous knowledge.
While the present study is new and promising, there are still many open questions to be addressed in future research. To account for human adaption, the learning control objective function may be designed to incorporate human-prosthesis performance feedback such as symmetry, user comfort level and satisfaction, and user energy consumption. To make the RL based control parameter tuning practical in real life, it should be further developed and tested in different use environments, such as from level ground walking to stair ascending or descending, and from walking to running.
The Smoothened receptor (SMO) belongs to the Class Frizzled of the G protein-coupled receptor (GPCR) superfamily, constituting a key component of the Hedgehog signalling pathway. Here we report the crystal structure of the multi-domain human SMO, bound and stabilized by a designed tool ligand TC114, using an X-ray free-electron laser source at 2.9 Å. The structure reveals a precise arrangement of three distinct domains: a seven-transmembrane helices domain (TMD), a hinge domain (HD) and an intact extracellular cysteine-rich domain (CRD). This architecture enables allosteric interactions between the domains that are important for ligand recognition and receptor activation. By combining the structural data, molecular dynamics simulation, and hydrogen-deuterium-exchange analysis, we demonstrate that transmembrane helix VI, extracellular loop 3 and the HD play a central role in transmitting the signal employing a unique GPCR activation mechanism, distinct from other multi-domain GPCRs.