Predicting Market Impact Costs Using Nonparametric Machine Learning Models

Market impact cost is the most significant portion of implicit transaction costs that can reduce the overall transaction cost, although it cannot be measured directly. In this paper, we employed the state-of-the-art nonparametric machine learning models: neural networks, Bayesian neural network, Gaussian process, and support vector regression, to predict market impact cost accurately and to provide the predictive model that is versatile in the number of variables. We collected a large amount of real single transaction data of US stock market from Bloomberg Terminal and generated three independent input variables. As a result, most nonparametric machine learning models outperformed a-state-of-the-art benchmark parametric model such as I-star model in four error measures. Although these models encounter certain difficulties in separating the permanent and temporary cost directly, nonparametric machine learning models can be good alternatives in reducing transaction costs by considerably improving in prediction performance.     

Cardiovascular Disorder Severity Detection Using Myocardial Anatomic Features Based Optimized Extreme Learning Machine Approach

ObjectivesThis study focuses on integration of anatomical left ventricle myocardium features and optimized extreme learning machine (ELM) for discrimination of subjects with normal, mild, moderate and severe abnormal ejection fraction (EF). The physiological alterations in myocardium have diagnostic relevance to the etiology of cardiovascular diseases (CVD) with reduced EF.Materials and MethodsThis assessment is carried out on cardiovascular magnetic resonance (CMR) images of 104 subjects available in Kaggle Second Annual Data Science Bowl. The Segment CMR framework is used to segment myocardium from cardiac MR images, and it is subdivided into 16 sectors. 86 clinically significant anatomical features are extracted and subjected to ELM framework. Regularization coefficient and hidden neurons influence the prediction accuracy of ELM. The optimal value for these parameters is achieved with the butterfly optimizer (BO). A comparative study of BOELM framework with different activation functions and feature set has been conducted.ResultsAmong the individual feature set, myocardial volume at ED gives a better classification accuracy of 83.3% compared to others. Further, the given BOELM framework is able to provide higher multi-class accuracy of 95.2% with the entire feature set than ELM. Better discrimination of healthy and moderate abnormal subjects is achieved than other sub groups.ConclusionThe combined anatomical sector wise myocardial features assisted BOELM is able to predict the severity levels of CVDs. Thus, this study supports the radiologists in the mass diagnosis of cardiac disorder.     

Estimation of Solvent Diffusion Coefficients Using Molecular Dynamics Simulations

Abstract

Molecular dynamics simulations have been used to investigate diffusion in two commonly used industrial solvents, toluene and tetrahydrofuran. Several different models for the solvents are compared (flexible vs. rigid, all-atom vs. united atom), and it is found that united atom and all-atom models of the solvents produce very different diffusion coefficients at the experimental density. This disagreement can be explained by the pressure dependence of the diffusion coefficient, which is found to vary in accord with the Chapman-Enskog result for hard spheres. It is recommended that force fields be parametrized carefully to produce reasonable pressures at the experimental densities, or that simulations be carried out at constant pressure, if they are to be used for the purposes of calculating transport coefficients.     

Enhancing Protein Adsorption Simulations by Using Accelerated Molecular Dynamics

The atomistic modeling of protein adsorption on surfaces is hampered by the different time scales of the simulation ( s) and experiment (up to hours), and the accordingly different ‘final’ adsorption conformations. We provide evidence that the method of accelerated molecular dynamics is an efficient tool to obtain equilibrated adsorption states. As a model system we study the adsorption of the protein BMP-2 on graphite in an explicit salt water environment. We demonstrate that due to the considerably improved sampling of conformational space, accelerated molecular dynamics allows to observe the complete unfolding and spreading of the protein on the hydrophobic graphite surface. This result is in agreement with the general finding of protein denaturation upon contact with hydrophobic surfaces.     

Structural Analysis of Human Lysozyme Using Molecular Dynamics Simulations

Abstract

In this study, various molecular dynamics simulations were conducted to investigate the effects of ethanol and temperature on the conformational changes of human lysozyme, which may lead insights into amyloidosis. The analyses of some important structural characteristics, such as backbone root-mean-square deviation, secondary structural stability, radius of gyration, accessible surface area, and hydrophobic contact of the hydrophobic core all show that ethanol tends to destabilize human lysozyme at high temperatures. It can be attributed to that higher temperatures result in the destruction of the native structure of this protein, leading to the exposure of the interior hydrophobic core. At this stage, ethanol plays a role to destroy this region by forming hydrophobic interactions between protein and solvent due to its lower polarity comparing to water. Such newly formed intermolecular interactions accelerate the unfolding of this protein, starting from the core between the a- and β-domains. Our results are in good agreement with the previous hypothesis suggesting that the distortion of the hydrophobic core at the α- and β-interface putatively results in the formation of the initial “seed” for amyloid fibril. Although the present results cannot directly be linked to fibril formation, they still provide valuable insights into amyloidosis of human lysozyme.     

A Hybrid Deep Learning Model for Predicting Molecular Subtypes of Human Breast Cancer Using Multimodal Data

BackgroundThe prediction of breast cancer subtypes plays a key role in the diagnosis and prognosis of breast cancer. In recent years, deep learning (DL) has shown good performance in the intelligent prediction of breast cancer subtypes. However, most of the traditional DL models use single modality data, which can just extract a few features, so it cannot establish a stable relationship between patient characteristics and breast cancer subtypes.DatasetWe used the TCGA-BRCA dataset as a sample set for molecular subtype prediction of breast cancer. It is a public dataset that can be obtained through the following link: https://portal.gdc.cancer.gov/projects/TCGA-BRCAMethodsIn this paper, a Hybrid DL model based on the multimodal data is proposed. We combine the patient's gene modality data with image modality data to construct a multimodal fusion framework. According to the different forms and states, we set up feature extraction networks respectively, and then we fuse the output of the two feature networks based on the idea of weighted linear aggregation. Finally, the fused features are used to predict breast cancer subtypes. In particular, we use the principal component analysis to reduce the dimensionality of high-dimensional data of gene modality and filter the data of image modality. Besides, we also improve the traditional feature extraction network to make it show better performance.ResultsThe results show that compared with the traditional DL model, the Hybrid DL model proposed in this paper is more accurate and efficient in predicting breast cancer subtypes. Our model achieved a prediction accuracy of 88.07% in 10 times of 10-fold cross-validation. We did a separate AUC test for each subtype, and the average AUC value obtained was 0.9427. In terms of subtype prediction accuracy, our model is about 7.45% higher than the previous average.     

Molecular Insights of Genetic Variation in Erianthus arundinaceus Populations Native to China

Background

E. arundinaceus (Retz.) Jeswiet is a warm-season, tall-growing perennial species native to much southern portion in China. The grass has been extensively used in sugarcane breeding and is recently targeted as a bioenergy feedstock crop. However, information on the genetic structure of the Chinese wild germplasm is limited. Knowledge of genetic variation within and among populations is essential for breeding new cultivars in the species. The major objective of this study was to quantify the magnitude of genetic variation among and within natural populations in China.

Methodology/Principal Findings

In this experiment, we analyzed genetic variation of 164 individuals of 18 populations collected from natural habitats in six Chinese provinces using 20 sequence-related amplified polymorphism (SRAP) primer pairs generating 277 polymorphic bands. Among and within the populations, the percentage of polymorphic bands (PPB) was 80.00% and 27.07%, genetic diversity (H_E) was 0.245 and 0.099, effective number of alleles (N_E) was 1.350 and 1.170, and Shannon''s information index (I) was 0.340 and 0.147, respectively. The populations were clustered into six groups exhibiting a high level of genetic differentiation, which was highly associated with geographic origins of respective germplasm populations, but was not significantly associated with geographic distances between the populations.

Conclusions/Significance

This is the first report indicating that large genetic variation exists in the Chinese E. arundinaceus germplasm based on the SRAP molecular marker analysis of native populations. The genetic structure of populations in the species has been substantially affected by geographic landforms and environments. The diverse collection will be highly valuable in genetic improvement in the species per se and likely in sugarcane.     

Molecular Insights from Conformational Ensembles via Machine Learning

Biomolecular simulations are intrinsically high dimensional and generate noisy data sets of ever-increasing size. Extracting important features from the data is crucial for understanding the biophysical properties of molecular processes, but remains a big challenge. Machine learning (ML) provides powerful dimensionality reduction tools. However, such methods are often criticized as resembling black boxes with limited human-interpretable insight. We use methods from supervised and unsupervised ML to efficiently create interpretable maps of important features from molecular simulations. We benchmark the performance of several methods, including neural networks, random forests, and principal component analysis, using a toy model with properties reminiscent of macromolecular behavior. We then analyze three diverse biological processes: conformational changes within the soluble protein calmodulin, ligand binding to a G protein-coupled receptor, and activation of an ion channel voltage-sensor domain, unraveling features critical for signal transduction, ligand binding, and voltage sensing. This work demonstrates the usefulness of ML in understanding biomolecular states and demystifying complex simulations.     

Molecular Dynamics Simulations of Zinc Ions in Water Using CHARMM

Molecular Dynamics simulations of a zinc ion with 123 and 525 TIP3P-water molecules were carried out with CHARMM using two different Lennard-Jones parameter sets for the Zn²⁺ ion. The results were compared to published experimental and simulation data. Good agreement was found for radial distribution functions, number of hydrogen bonds, and diffusion coefficients. Experimental radial distribution functions were better reproduced by the original CHARMM22 parameter set than by the parameter set modified by Stote and Karplus. Diffusion coefficients were found to depend on the system size rather than on the parameter set used and were better reproduced by the larger systems. The divalent zinc ion exerts a strong influence on its hydration shell as indicated by the high first peak of the radial distribution function. Water molecules in the vicinity of the zinc ion show a slight deformation of the O-H bond length and of the H-O-H bond angle as compared to pure water. No water molecules from the first hydration shell were exchanged during 1 ns of MD simulation.Electronic Supplementary Material available.     

A Method to Predict Blood-Brain Barrier Permeability of Drug-Like Compounds Using Molecular Dynamics Simulations

The blood-brain barrier (BBB) is formed by specialized tight junctions between endothelial cells that line brain capillaries to create a highly selective barrier between the brain and the rest of the body. A major problem to overcome in drug design is the ability of the compound in question to cross the BBB. Neuroactive drugs are required to cross the BBB to function. Conversely, drugs that target other parts of the body ideally should not cross the BBB to avoid possible psychotropic side effects. Thus, the task of predicting the BBB permeability of new compounds is of great importance. Two gold-standard experimental measures of BBB permeability are logBB (the concentration of drug in the brain divided by concentration in the blood) and logPS (permeability surface-area product). Both methods are time-consuming and expensive, and although logPS is considered the more informative measure, it is lower throughput and more resource intensive. With continual increases in computer power and improvements in molecular simulations, in silico methods may provide viable alternatives. Computational predictions of these two parameters for a sample of 12 small molecule compounds were performed. The potential of mean force for each compound through a 1,2-dioleoyl-sn-glycero-3-phosphocholine bilayer is determined by molecular dynamics simulations. This system setup is often used as a simple BBB mimetic. Additionally, one-dimensional position-dependent diffusion coefficients are calculated from the molecular dynamics trajectories. The diffusion coefficient is combined with the free energy landscape to calculate the effective permeability (P_eff) for each sample compound. The relative values of these permeabilities are compared to experimentally determined logBB and logPS values. Our computational predictions correlate remarkably well with both logBB (R² = 0.94) and logPS (R² = 0.90). Thus, we have demonstrated that this approach may have the potential to provide reliable, quantitatively predictive BBB permeability, using a relatively quick, inexpensive method.     

A Method to Predict Blood-Brain Barrier Permeability of Drug-Like Compounds Using Molecular Dynamics Simulations

The blood-brain barrier (BBB) is formed by specialized tight junctions between endothelial cells that line brain capillaries to create a highly selective barrier between the brain and the rest of the body. A major problem to overcome in drug design is the ability of the compound in question to cross the BBB. Neuroactive drugs are required to cross the BBB to function. Conversely, drugs that target other parts of the body ideally should not cross the BBB to avoid possible psychotropic side effects. Thus, the task of predicting the BBB permeability of new compounds is of great importance. Two gold-standard experimental measures of BBB permeability are logBB (the concentration of drug in the brain divided by concentration in the blood) and logPS (permeability surface-area product). Both methods are time-consuming and expensive, and although logPS is considered the more informative measure, it is lower throughput and more resource intensive. With continual increases in computer power and improvements in molecular simulations, in silico methods may provide viable alternatives. Computational predictions of these two parameters for a sample of 12 small molecule compounds were performed. The potential of mean force for each compound through a 1,2-dioleoyl-sn-glycero-3-phosphocholine bilayer is determined by molecular dynamics simulations. This system setup is often used as a simple BBB mimetic. Additionally, one-dimensional position-dependent diffusion coefficients are calculated from the molecular dynamics trajectories. The diffusion coefficient is combined with the free energy landscape to calculate the effective permeability (P_eff) for each sample compound. The relative values of these permeabilities are compared to experimentally determined logBB and logPS values. Our computational predictions correlate remarkably well with both logBB (R² = 0.94) and logPS (R² = 0.90). Thus, we have demonstrated that this approach may have the potential to provide reliable, quantitatively predictive BBB permeability, using a relatively quick, inexpensive method.     

Refined Genetic Algorithm Simulations to Model Proteins

The advent of completely sequenced genomes is leading to an unprecedented growth of sequence information while adequate structure information is often lacking. Genetic algorithm simulations have been refined and applied as a helpful tool for this question. Modified strategies are tested first on simple lattice protein models. This includes consideration of entropy (protein adjacent water shell) and improved search strategies (pioneer search +14%, systematic recombination +50% in search efficiency). Next, extension to grid free simulations of proteins in full main chain representation is examined. Our protein main chain simulations are further refined by independent criteria such as fitness per residue to judge predicted structures obtained at the end of a simulation. Protein families and protein interactions predicted from the complete H. pylori genomic sequence demonstrate how the full main chain simulations are then applied to model new protein sequences and protein families apparent from genome analysis.     

Using Molecular Dynamics Simulations on Crambin to Evaluate the Suitability of Different Continuum Dielectric and Hydrogen Atom Models for Protein Simulations

Abstract

Molecular dynamics simulations of enzymes with enough explicit waters of solvation to realistically account for solute-solvent interactions can burden the computational resources required to perform the simulation by more than two orders of magnitude. Since enzyme simulations even with an implicit solvation model can be imposing for a supercomputer, it is important to assess the suitability of different continuum dielectric models for protein simulations. A series of 100-picosecond molecular dynamics simulations were performed on the X-ray crystal structure of the protein crambin to examine how well computed structures, obtained using seven continuum dielectric and two hydrogen atom models, agreed with the X-ray structure. The best level of agreement between computed and experimental structures was obtained using a constant dielectric of 2 and the all-hydrogen model. Continuum dielectric models of 1,1^*r, and 2^*r also led to computed structures in reasonably good agreement with the X-ray structure. In all cases, the all-hydrogen model gave better agreement than the united atom model, although, in one case, the difference was not significant. Dielectric models of 4, 80, and 4^*r with either hydrogen model yielded significantly poorer fits. It is especially noteworthy that the observed trends did not semiquantitatively converge until about 50 picoseconds into the simulations, suggesting that validation studies for protein calculations based on energy minimizations or short simulations should be viewed with caution.     

Probing the Flexibility of Tropomyosin and Its Binding to Filamentous Actin Using Molecular Dynamics Simulations

Tropomyosin (Tm) is a coiled-coil protein that binds to filamentous actin (F-actin) and regulates its interactions with actin-binding proteins like myosin by moving between three positions on F-actin (the blocked, closed, and open positions). To elucidate the molecular details of Tm flexibility in relation to its binding to F-actin, we conducted extensive molecular dynamics simulations for both Tm alone and Tm-F-actin complex in the presence of explicit solvent (total simulation time >400 ns). Based on the simulations, we systematically analyzed the local flexibility of the Tm coiled coil using multiple parameters. We found a good correlation between the regions with high local flexibility and a number of destabilizing regions in Tm, including six clusters of core alanines. Despite the stabilization by F-actin binding, the distribution of local flexibility in Tm is largely unchanged in the absence and presence of F-actin. Our simulations showed variable fluctuations of individual Tm periods from the closed position toward the open position. In addition, we performed Tm-F-actin binding calculations based on the simulation trajectories, which support the importance of Tm flexibility to Tm-F-actin binding. We identified key residues of Tm involved in its dynamic interactions with F-actin, many of which have been found in recent mutational studies to be functionally important, and the rest of which will make promising targets for future mutational experiments.     

Probing the Flexibility of Tropomyosin and Its Binding to Filamentous Actin Using Molecular Dynamics Simulations

Tropomyosin (Tm) is a coiled-coil protein that binds to filamentous actin (F-actin) and regulates its interactions with actin-binding proteins like myosin by moving between three positions on F-actin (the blocked, closed, and open positions). To elucidate the molecular details of Tm flexibility in relation to its binding to F-actin, we conducted extensive molecular dynamics simulations for both Tm alone and Tm-F-actin complex in the presence of explicit solvent (total simulation time >400 ns). Based on the simulations, we systematically analyzed the local flexibility of the Tm coiled coil using multiple parameters. We found a good correlation between the regions with high local flexibility and a number of destabilizing regions in Tm, including six clusters of core alanines. Despite the stabilization by F-actin binding, the distribution of local flexibility in Tm is largely unchanged in the absence and presence of F-actin. Our simulations showed variable fluctuations of individual Tm periods from the closed position toward the open position. In addition, we performed Tm-F-actin binding calculations based on the simulation trajectories, which support the importance of Tm flexibility to Tm-F-actin binding. We identified key residues of Tm involved in its dynamic interactions with F-actin, many of which have been found in recent mutational studies to be functionally important, and the rest of which will make promising targets for future mutational experiments.     

Molecular Dynamics Simulations of a DMPC Bilayer Using Nonadditive Interaction Models

We present a polarizable force field based on the charge-equilibration formalism for molecular dynamics simulations of phospholipid bilayers. We discuss refinement of headgroup dihedral potential parameters to reproduce ab initio conformational energies of dimethylphosphate calculated at the MP2/cc-pVTZ level of theory. We also address the refinement of electrostatic and Lennard-Jones (van der Waals) parameters to reproduce ab initio polarizabilities and water interaction energies of dimethylphosphate and tetramethylammonium. We present results of molecular dynamics simulations of a solvated dimyristoylphosphatidylcholine bilayer using this polarizable force field as well as the nonpolarizable, fixed-charge CHARMM27 and CHARMM27r force fields for comparison. Calculated atomic and electron-density profiles, deuterium order parameters, and headgroup orientations are found to be consistent with previous simulations and with experiment. Polarizable interaction models for solvent and lipid exhibit greater water penetration into the lipid interior; this is due to the variation of water molecular dipole moment from a bulk value of 2.6 Debye to a value of 1.9 Debye in the membrane interior. The reduction in the electrostatic component of the desolvation free-energy penalty allows for greater water density. The surface dipole potential predicted by the polarizable model is 0.95 V compared to the value of 0.8 V based on nonpolarizable force-field calculations. Effects of inclusion of explicit polarization are discussed in relation to water dipole moment and varying charge distributions. Dielectric permittivity profiles for polarizable and nonpolarizable interactions exhibit subtle differences arising from the nature of the individual component parameterizations; for the polarizable force field, we obtain a bulk dielectric permittivity of 79, whereas the nonpolarizable force field plateaus at 97 (the value for pure TIP3P water). In the membrane interior, both models predict unit permittivities, with the polarizable models contributing from one to two more units due to the optical dielectric (high-frequency dipole fluctuations). This contribution is a step toward the continuing development of a CHARMM (Chemistry at Harvard Molecular Mechanics) polarizable force field for simulations of biomacromolecular systems.