Background: Machine learning (ML) has made a significant impact in medicine and cancer research; however, its impact in these areas has been undeniably slower and more limited than in other application domains. A major reason for this has been the lack of availability of patient data to the broader ML research community, in large part due to patient privacy protection concerns. High-quality, realistic, synthetic datasets can be leveraged to accelerate methodological developments in medicine. By and large, medical data is high dimensional and often categorical. These characteristics pose multiple modeling challenges. Methods: In this paper, we evaluate three classes of synthetic data generation approaches; probabilistic models, classification-based imputation models, and generative adversarial neural networks. Metrics for evaluating the quality of the generated synthetic datasets are presented and discussed. Results: While the results and discussions are broadly applicable to medical data, for demonstration purposes we generate synthetic datasets for cancer based on the publicly available cancer registry data from the Surveillance Epidemiology and End Results (SEER) program. Specifically, our cohort consists of breast, respiratory, and non-solid cancer cases diagnosed between 2010 and 2015, which includes over 360,000 individual cases. Conclusions: We discuss the trade-offs of the different methods and metrics, providing guidance on considerations for the generation and usage of medical synthetic data.

SUMMARY Dynamin is a 100 kDa GTPase that organizes into helical assemblies at the base of nascent clathrin-coated vesicles. Formation of these oligomers stimulates the intrinsic GTPase activity of dynamin, which is necessary for efficient membrane fission during endocytosis. Recent evidence suggests that the transition-state of dynamin's GTP hydrolysis reaction serves as a key determinant of productive fission. Here we present the structure of a transition-state-defective dynamin mutant, K44A, trapped in a pre-fission state, at 12.5 Å resolution. This structure constricts to 3.7 nm, reaching the theoretical limit required for spontaneous membrane fission. Computational docking indicates that the ground state conformation of the dynamin polymer is sufficient to achieve this super-constricted pre-fission state and reveals how a 2-start helical symmetry promotes the most efficient packing of dynamin tetramers around the membrane neck. These data suggest a new model for the assembly and regulation of the minimal dynamin fission machine.

Numerous vesiculation processes throughout the eukaryotic cell are dependent on the protein dynamin, a large GTPase that constricts lipid bilayers. We have combined X-ray crystallography and cryo-electron microscopy (cryo-EM) data to generate a coherent model of dynamin-mediated membrane constriction. GTPase and pleckstrin homology domains of dynamin were fit to cryo-EM structures of human dynamin helices bound to lipid in nonconstricted and constricted states. Proteolysis and immunogold labeling experiments confirm the topology of dynamin domains predicted from the helical arrays. Based on the fitting, an observed twisting motion of the GTPase, middle, and GTPase effector domains coincides with conformational changes determined by cryo-EM. We propose a corkscrew model for dynamin constriction based on these motions and predict regions of sequence important for dynamin function as potential targets for future mutagenic and structural studies.