Refine
Document Type
- Doctoral Thesis (4)
Language
- English (4)
Has Fulltext
- yes (4)
Is part of the Bibliography
- no (4)
Keywords
- deep learning (1)
- denoising (1)
- receptor tyrosine kinases (1)
- super-resolution microscopy (1)
- virtual labeling (1)
Institute
The diffusive behavior of macromolecules in solution is a key factor in the kinetics of macromolecular binding and assembly, and in the theoretical description of many experiments. Experiments on high-density protein solutions have found that a slow down of the diffusion dynamics is larger than expected from colloidal theory for non-interaction hard-spheres. It has also been shown that the rotational diffusion anisotropy in high-density protein solutions is larger than in dilute ones. High-density protein solutions are a complex fluid that is different from the neat fluid assumption used in the hydrodynamic theory. It is therefore important to have methods to accurately calculate the translational and rotational diffusion tensor from simulations as well as simulation algorithms to explore high-density solutions.
Simulations provide a powerful tool to study diffusion in complex fluids. They can be used to study the macroscopic and microscopic effects of complex fluids on the diffusive behavior. There has been already a lot of work done to accurately simulate diffusion and to determine the diffusion coefficients from simulations.
The translational diffusion of molecules in simple and complex liquids can be determined with high accuracy from simulations. This is not yet the case for rotational diffusion. Existing algorithms to calculate the rotational diffusion coefficients from simulations make assumptions about the shape of the protein or only work at short times. For the simulation of diffusive behavior of macromolecules two options exist today. An all-atom integrator with explicit solvent molecules or coarse-grained (CG) simulations with an implicit solvent. CG simulations of dynamic behavior with implicit solvent are also called Brownian dynamics (BD) simulations. For the CG simulations the Ermak-McCammon algorithm is often used to solve the underlying Langevin equation. The algorithm is an extension of the Euler-Maruyama integrator to include translation and rotation in three dimensions. This algorithm only correctly reproduces the equilibrium probability for short time-steps and the error depends linearly on the time-step. It has been shown that Monte Carlo based algorithms can produce BD for translational dynamics, when appropriately parametrized. The advantage of Monte Carlo based algorithm is that they will reproduce the correct equilibrium distribution independent of the chosen time-step. This in return allows choosing larger time-steps in simulations. The aim of this thesis is to develop novel´methods to accurately determine the rotational diffusion coefficient from simulations and extend existing Monte Carlo algorithms to include rotational dynamics.
The first project addresses the question of how to accurately determine the rotational diffusion coefficients from simulations. We develop a quaternion based method to calculate the rotational diffusion tensor from simulations and a theory for the effects of periodic boundary conditions (PBC) on the rotational diffusion coefficient in simulations.
Our method for calculating rotational diffusion coefficients is based on the quaternion covariances from Favro for a freely rotating rigid molecule. The covariances as formulated by Favro are only valid in the principal coordinate system (PCS) of the rotation diffusion tensor. The covariances can be generalized for an arbitrary reference coordinate system (RCS), i.e., a simulation, given the principle axes of the rotational diffusion tensor in the RCS. We show that no prior knowledge of the diffusion tensor and its principal axes is required to calculate the generalized covariances from simulations using common root-mean-square distance (RMSD) procedures. We develop two methods to fit the covariances calculated from simulations to our generalized equations to fit the rotational diffusion tensor. In the first method we minimize the sum of the squared error deviations between model and simulation data. For this six dimensional optimization we use a simulated annealing algorithm. Alternatively the rotational diffusion tensor can also be determined from a eigenvalue decomposition of covariance after integration. To minimize the effects of sampling noise in the integration we first apply a Laplace-transformation to smooth the covariances at large times. For ideal sampling the resulting rotational diffusion coefficient should be independent of the value of the Laplace variable. In practice, however, the best results are achieved using a value close to the inverse autocorrelation time of the rotational motion.
...
Cryo-electron tomography (CET) is a unique technique to visualize biological objects under near-to-native conditions at near-atomic resolution. CET provides three-dimensional (3D) snapshots of the cellular proteome, in which the spatial relations between macromolecular complexes in their near native cellular context can be explored. Due to the limitation of the electron dose applicable on biological samples, the achievable resolution of a tomogram is restricted to a few nanometers, higher resolution can be achieved by averaging of structures occurring in multiples. For this purpose, computational techniques such as template matching, sub-tomogram averaging and classification are essential for a meaningful processing of CET data.
This thesis introduces the techniques of template matching and sub-tomogram averaging and their applications on real biological data sets. Subsequently, the problem of reference bias, which restricts the applicability of those techniques, is addressed. Two methods that estimate the reference bias in Fourier and real space are demonstrated. The real space method, which we have named the “M-free” score, provides a reliable estimation of the reference bias, which gives access to the reliability of the template matching or sub-tomogram averaging process. Thus, the “M-free” score makes those approaches more applicable to structural biology. Furthermore, a classification algorithm based on Neural Networks (NN) called “KerDenSOM3D” is introduced, which is implemented in 3D and compensates for the missing-wedge. This approach helps extracting different structural states of macromolecular complexes or increasing the class purity of data sets by eliminating outliers. A comprehensive comparison with other classification methods shows superior performance of KerDenSOM3D.
As its fundamental function, the brain processes and transmits information using populations of interconnected nerve cells alias neurons. The communication between these neurons occurs via discrete electric impulses called spikes. A core challenge in neuroscience has been to quantify how much information about relevant stimuli or signals a neuron transports in its spike sequences, or spike trains. The recently introduced correlation method allows to determine this so-called mutual information in terms of a neuron’s temporal spike correlations under certain stationarity assumptions. Based on the correlation method, I address several open questions regarding neural information encoding in the cortex.
In the first part (chapter 2), I investigate the role of temporal spike correlations for neural information transmission. Temporal correlations in neuronal spike trains diminish independence in the information that is transmitted by the different spikes and hence introduce redundancy to stimulus encoding. However, exact methods to describe how such spike correlations impact information transmission quantitatively have been lacking. Here, I provide a general measure for the information carried by spike trains of neurons with correlated rate modulations only, neglecting other spike correlations, and use it to investigate the effect of rate correlations on encoding redundancy. I derive it analytically by calculating the mutual information between a time correlated, rate-modulating signal and the resulting spikes of Poisson neurons. Whereas this information is determined by spike autocorrelations only, the redundancy in information encoding due to rate correlations depends on both the distribution and the autocorrelation of the rate histogram. I further demonstrate that, at very small signal strengths, the information carried by rate correlated spikes becomes identical to that of independent spikes, in effect measuring the rate modulation depth. In contrast, a vanishing signal correlation time maximizes information transmission but does not generally yield the information of independent spikes.
In the second part (chapter 3), I analyze the information transmission capabilities of two particular schemes of encoding stimuli in the synaptic inputs using integrate-and-fire neuron models. Specifically, I calculate the exact information contained in spike trains about signals which modulate either the mean or the variance of the somatic currents in neurons, as is observed experimentally. I show that the information content about mean modulating signals is generally substantially larger than about variance modulating signals for biological parameters. This result provides evidence, by means of exact calculations of the mutual information, against the potential benefit of variance encoding that had been suggested previously.
Another analysis reveals that higher information transmission is generally associated with a larger proportion of nonlinear signal encoding. Moreover, I show that a combination of signal-dependent mean and variance modulations of the input current can synergistically benefit information transmission through a nonlinear coupling of both channels. On a more general level, I identify what was previously considered an upper bound as the exact, full mutual information. Furthermore, by analyzing the statistics of the spike train Fourier coefficients, I identify the means of the Fourier coefficients as information-carrying features.
Overall, this work contributes answers to central questions of theoretical neuroscience concerning the neural code and neural information transmission. It sheds light on the role of signal-induced temporal correlations for neural coding by providing insight into how signal features shape redundancy and by establishing mathematical links between existing methods and providing new insights into the spike train statistics in stationary situations. Moreover, I determine what fraction of the mutual information is linearly decodable for two specific signal encoding schemes.
Fluorescence microscopy has significantly impacted our understanding of cell biology. The extension of diffraction-unlimited super-resolution microscopy opened an observation window that allows for the scrutiny of cellular organization at a molecular level. The non-invasive nature of visible light in super-resolution microscopy methods renders them suitable for observations in living cells and organisms. Building upon these advancements, a promising synergy between super-resolution fluorescence microscopy and deep learning becomes evident, extending the capabilities of the imaging methods. Tasks such as image modality translation, restoration, single-molecule fitting, virtual labeling, spectral demixing, and molecular counting, are enabled with high precision. The techniques explored in this thesis address three critical facets in advanced microscopy, namely the reduction in image acquisition time, saving photon budget during measurement, and increasing the multiplexing capability. Furthermore, descriptors of protein distributions and their motion on cell membranes were developed.