Slavica Jonic

Three machine learning techniques for automatic determination of rules to control locomotion

Automatic prediction of gait events (e,g,, heel contact, flat foot, initiation of the swing, etc.) and corresponding profiles of the activations of muscles is important for real-time control of locomotion. This paper presents three supervised machine learning (ML) techniques for prediction of the activation patterns of muscles and sensory data, based on the history of sensory data, for walking assisted by a functional electrical stimulation (FES). Those MLs are: (1) a multilayer perceptron with Levenberg-Marquardt modification of backpropagation learning algorithm; (2) an adaptive-network-based fuzzy inference system (ANFIS); and (3) a combination of an entropy minimization type of inductive learning (IL) technique and a radial basis function (RBF) type of artificial neural network with orthogonal least squares learning algorithm. Here we show the prediction of the activation of the knee flexor muscles and the knee joint angle for seven consecutive strides based on the history of the knee joint angle and the ground reaction forces. The data used for training and testing of MLs was obtained from a simulation of walking assisted with an FES system. The ability of generating rules for an FES controller was selected as the most important criterion when comparing the MLs. Other criteria such as generalization of results, computational complexity, and learning rate mere also considered. The minimal number of rules and the most explicit and comprehensible rules were obtained by ANFIS. The best generalization was obtained by the IL and RBF network.

Optimal control of walking with functional electrical stimulation: a computer simulation study

Bipedal locomotion was simulated to generate a pattern of activating muscles for walking using electrical stimulation in persons with spinal cord injury (SCI) or stroke. The simulation presented in this study starts from a model of the body determined with user-specific parameters, individualized with respect to the lengths, masses, inertia, muscle and joint properties. The trajectory used for simulation was recorded from an able-bodied subject while walking with ankle-foot orthoses. A discrete mathematical model and dynamic programming were used to determine the optimal control. A cost function was selected as the sum of the squares of the tracking errors from the desired trajectories, and the weighted sum of the squares of agonist and antagonist activations of the muscle groups acting around the hip and knee joints. The aim of the simulation was to study plausible trajectories keeping in mind the limitations imposed by the spinal cord injury or stroke (e.g., spasticity, decreased range of movements in some joints, limited strength of paralyzed, externally activated muscles). If the muscles were capable of generating the movements required and the trajectory was achieved, then the simulation provided two kinds of information: 1) timing of the onset and offset of muscle activations with respect to the various gait events and 2) patterns of activation with respect to the maximum activation. These results are important for synthesizing a rule-based controller.

Full-open and closed CFTR channels, with lateral tunnels from the cytoplasm and an alternative position of the F508 region, as revealed by molecular dynamics

In absence of experimental 3D structures, several homology models, based on ABC exporter 3D structures, have provided significant insights into the molecular mechanisms underlying the function of the cystic fibrosis transmembrane conductance regulator (CFTR) protein, a chloride channel whose defects are associated with cystic fibrosis (CF). Until now, these models, however, did not furnished much insights into the continuous way that ions could follow from the cytosol to the extracellular milieu in the open form of the channel. Here, we have built a refined model of CFTR, based on the outward-facing Sav1866 experimental 3D structure and integrating the evolutionary and structural information available today. Molecular dynamics simulations revealed significant conformational changes, resulting in a full-open channel, accessible from the cytosol through lateral tunnels displayed in the long intracellular loops (ICLs). At the same time, the region of nucleotide-binding domain 1 in contact with one of the ICLs and carrying amino acid F508, the deletion of which is the most common CF-causing mutation, was found to adopt an alternative but stable position. Then, in a second step, this first stable full-open conformation evolved toward another stable state, in which only a limited displacement of the upper part of the transmembrane helices leads to a closure of the channel, in a conformation very close to that adopted by the Atm1 ABC exporter, in an inward-facing conformation. These models, supported by experimental data, provide significant new insights into the CFTR structure–function relationships and into the possible impact of CF-causing mutations.

The Structure of Phosphorylase Kinase Holoenzyme at 9.9 Å Resolution and Location of the Catalytic Subunit and the Substrate Glycogen Phosphorylase

Summary Phosphorylase kinase (PhK) coordinates hormonal and neuronal signals to initiate the breakdown of glycogen. The enzyme catalyzes the phosphorylation of inactive glycogen phosphorylase b (GPb), resulting in the formation of active glycogen phosphorylase a. We present a 9.9 Å resolution structure of PhK heterotetramer (αβγδ)4 determined by cryo-electron microscopy single-particle reconstruction. The enzyme has a butterfly-like shape comprising two lobes with 222 symmetry. This three-dimensional structure has allowed us to dock the catalytic γ subunit to the PhK holoenzyme at a location that is toward the ends of the lobes. We have also determined the structure of PhK decorated with GPb at 18 Å resolution, which shows the location of the substrate near the kinase subunit. The PhK preparation contained a number of smaller particles whose structure at 9.8 Å resolution was consistent with a proteolysed activated form of PhK that had lost the α subunits and possibly the γ subunits.

Protein structure determination by electron cryo-microscopy

Transmission electron cryo-microscopy (cryoEM) is a versatile tool in the structural analysis of proteins and biological macromolecular assemblies. In this review, we present a brief survey of the methods used in cryoEM, and their current developments. These latest advances provide exciting opportunities for the three-dimensional structural determination of macromolecular complexes that are either too large or too heterogeneous to be investigated by conventional X-ray crystallography or nuclear magnetic resonance (NMR). The endeavour of understanding the function of protein or macromolecular complex is often helped by combining data from electron microscopy and X-ray crystallography. We will thus provide a brief overview of the computational techniques involved in combining data from different techniques for the interpretation of the EM structure.

Comparison of single-particle analysis and electron tomography approaches: an overview

Three‐dimensional structure of a wide range of biological specimens can be computed from images collected by transmission electron microscopy. This information integrated with structural data obtained with other techniques (e.g., X‐ray crystallography) helps structural biologists to understand the function of macromolecular complexes and organelles within cells. In this paper, we compare two three‐dimensional transmission electron microscopy techniques that are becoming more and more related (at the image acquisition level as well as the image processing one): electron tomography and single‐particle analysis. The first one is currently used to elucidate the three‐dimensional structure of cellular components or smaller entire cells, whereas the second one has been traditionally applied to structural studies of macromolecules and macromolecular complexes. Also, we discuss possibilities for their integration with other structural biology techniques for an integrative study of living matter from proteins to whole cells.

The Subnanometer Resolution Structure of the Glutamate Synthase 1.2-Mda Hexamer by Cryoelectron Microscopy and its Oligomerization Behavior in Solution: Functional Implications.

The three-dimensional structure of the hexameric (αβ)6 1.2-MDa complex formed by glutamate synthase has been determined at subnanometric resolution by combining cryoelectron microscopy, small angle x-ray scattering, and molecular modeling, providing for the first time a molecular model of this complex iron-sulfur flavoprotein. In the hexameric species, interprotomeric α-α and α-β contacts are mediated by the C-terminal domain of the α subunit, which is based on a β helical fold so far unique to glutamate synthases. The αβ protomer extracted from the hexameric model is fully consistent with it being the minimal catalytically active form of the enzyme. The structure clarifies the electron transfer pathway from the FAD cofactor on the β subunit, to the FMN on the α subunit, through the low potential [4Fe-4S]1+/2+ centers on the β subunit and the [3Fe-4S]0/1+ cluster on the α subunit. The (αβ)6 hexamer exhibits a concentration-dependent equilibrium with αβ monomers and (αβ)2 dimers, in solution, the hexamer being destabilized by high ionic strength and, to a lower extent, by the reaction product NADP+. Hexamerization seems to decrease the catalytic efficiency of the αβ protomer only 3-fold by increasing the Km values measured for l-Gln and 2-OG. However, it cannot be ruled out that the (αβ)6 hexamer acts as a scaffold for the assembly of multienzymatic complexes of nitrogen metabolism or that it provides a means to regulate the activity of the enzyme through an as yet unknown ligand.

FASTDEF: Fast defocus and astigmatism estimation for high-throughput transmission electron microscopy

In this work we present a fast and automated algorithm for estimating the contrast transfer function (CTF) of a transmission electron microscope. The approach is very suitable for High Throughput work because: (a) it does not require any initial defocus estimation, (b) it is almost an order of magnitude faster than existing approaches, (c) it opens the way to well-defined extensions to the estimation of higher order aberrations, at the same time that provides defocus and astigmatism estimations comparable in accuracy to well established methods, such as Xmipp and CTFFIND3 approaches. The new algorithm is based on obtaining the wrapped modulating phase of the power spectra density pattern by the use of a quadrature filter. This phase is further unwrapped in order to obtain the continuous and smooth absolute phase map; then a Zernike polynomial fitting is performed and the defocus and astigmatism parameters are determined. While the method does not require an initial estimation of the defocus parameters or any non-linear optimization procedure, these approaches can be used if further refinement is desired. Results of the CTF estimation method are presented for standard negative stained images, cryo-electron microscopy images in the absence of carbon support, as well as micrographs with only ice. Additionally, we have also tested the proposed method with micrographs acquired from tilted and untilted samples, obtaining good results. The algorithm is freely available as a part of the Xmipp package [http://xmipp.cnb.csic.es].

3D electron microscopy of biological nanomachines: principles and applications

Transmission electron microscopy is a powerful technique for studying the three-dimensional (3D) structure of a wide range of biological specimens. Knowledge of this structure is crucial for fully understanding complex relationships among macromolecular complexes and organelles in living cells. In this paper, we present the principles and main application domains of 3D transmission electron microscopy in structural biology. Moreover, we survey current developments needed in this field, and discuss the close relationship of 3D transmission electron microscopy with other experimental techniques aimed at obtaining structural and dynamical information from the scale of whole living cells to atomic structure of macromolecular complexes.

3D mapping of glycogenosis-causing mutations in the large regulatory alpha subunit of phosphorylase kinase

Mutations in the liver isoform of the Phosphorylase Kinase (PhK) alpha subunit (PHKA2 gene) cause X-linked liver glycogenosis (XLG), the most frequent type of PhK deficiency (glycogen-storage disease type IX). XLG patients can be divided in two subgroups, with similar clinical features but different activity of PhK (decreased in liver and blood cells for XLG-I and low in liver but normal or enhanced in blood cells for XLG-II). Here, we show that the PHKA2 missense mutations and small in-frame deletions/insertions are concentrated into two domains of the protein, which were recently described. In the N-terminal glucoamylase domain, mutations (principally leading to XLG-II) are clustered within the predicted glycoside-binding site, suggesting that they may have a direct impact on a possible hydrolytic activity of the PhK alpha subunit, which remains to be demonstrated. In the C-terminal calcineurin B-like domain (domain D), mutations (principally leading to XLG-I) are clustered in a region predicted to interact with the regulatory region of the PhK catalytic subunit and in a region covering this interaction site. Altogether, these results show that PHKA2 missense mutations or small in-frame deletions/insertions may have a direct impact on the PhK alpha functions and provide a framework for further experimental investigation.