Geometry Descriptors Research Articles

Machine learning-assisted amidase-catalytic enantioselectivity prediction and rational design of variants for improving enantioselectivity

Biocatalysis is an attractive approach for the synthesis of chiral pharmaceuticals and fine chemicals, but assessing and/or improving the enantioselectivity of biocatalyst towards target substrates is often time and resource intensive. Although machine learning has been used to reveal the underlying relationship between protein sequences and biocatalytic enantioselectivity, the establishment of substrate fitness space is usually disregarded by chemists and is still a challenge. Using 240 datasets collected in our previous works, we adopt chemistry and geometry descriptors and build random forest classification models for predicting the enantioselectivity of amidase towards new substrates. We further propose a heuristic strategy based on these models, by which the rational protein engineering can be efficiently performed to synthesize chiral compounds with higher ee values, and the optimized variant results in a 53-fold higher E-value comparing to the wild-type amidase. This data-driven methodology is expected to broaden the application of machine learning in biocatalysis research.

Mapping stress inside living cells by atomic force microscopy in response to environmental stimuli

ABSTRACT The response of cells to environmental stimuli, under either physiological or pathological conditions, plays a key role in determining cell fate toward either adaptive survival or controlled death. The efficiency of such a feedback mechanism is closely related to the most challenging human diseases, including cancer. Since cellular responses are implemented through physical forces exerted on intracellular components, more detailed knowledge of force distribution through modern imaging techniques is needed to ensure a mechanistic understanding of these forces. In this work, we mapped these intracellular forces at a whole-cell scale and with submicron resolution to correlate intracellular force distribution to the cytoskeletal structures. Furthermore, we visualized dynamic mechanical responses of the cells adapting to environmental modulations in situ. Such task was achieved by using an informatics-assisted atomic force microscope (AFM) indentation technique where a key step was Markov-chain Monte Carlo optimization to search for both the models used to fit indentation force–displacement curves and probe geometry descriptors. We demonstrated force dynamics within cytoskeleton, as well as nucleoskeleton in living cells which were subjected to mechanical state modulation: myosin motor inhibition, micro-compression stimulation and geometrical confinement manipulation. Our results highlight the alteration in the intracellular prestress to attenuate environmental stimuli; to involve in cellular survival against mechanical signal-initiated death during cancer growth and metastasis; and to initiate cell migration.

A ML framework to predict permeability of highly porous media based on PSD

Using machine learning (ML) method to predict permeability of porous media has shown great potential in recent years. A current problem is the lack of effective models to account for highly porous media with dilated pores. This study includes (1) generation of media (porosity = 0.8) via a Boolean process, (2) the pore size distribution (PSD) control by using different groups of homogeneous packed spherical particles (3) PSD data obtainment using the spherical contact distribution model (4) computation of the permeability via LBM simulations, (4) training of artificial neuron network (ANN) and (5) analysis of the model. It is found that the PSD could outperform the previous geometry descriptors as an input of ML framework to deal with highly porous structures with different fractions of dilated pores, however there is still room for precision enhancement.

Analysis of the Schwarz Domain Decomposition Method for the Conductor-like Screening Continuum Model

We study the Schwarz overlapping domain decomposition method applied to the Poisson problem on a special family of domains, which by construction consist of a union of a large number of fixed-size subdomains. These domains are motivated by applications in computational chemistry where the subdomains consist of van der Waals balls. As is usual in the theory of domain decomposition methods, the rate of convergence of the Schwarz method is related to a stable subspace decomposition. We derive such a stable decomposition for this family of domains and analyze how the stability "constant" depends on relevant geometric properties of the domain. For this, we introduce new descriptors that are used to formalize the geometry for the family of domains. We show how, for an increasing number of subdomains, the rate of convergence of the Schwarz method depends on specific local geometry descriptors and on one global geometry descriptor. The analysis also naturally provides lower bounds in terms of the descriptors for the smallest eigenvalue of the Laplace eigenvalue problem for this family of domains.

Using deep learning to extract novel and quantitative imaging features from perfluoropropane contrast, sulphur hexafluoride contrast and non-contrast stress echocardiography images

Abstract Background Stress echocardiography has become established as the most widely applied non-invasive imaging test for diagnosis of coronary artery disease within the UK. However, stress echocardiography has been substantially qualitative, rather than quantitative, based on visual wall motion assessment. For the first time, we have identified and validated quantitative descriptors of cardiac geometry and motion, extracted from ultrasound images acquired using contrast agents in an automated way. Purpose To establish whether these novel imaging features can be generated in an automated, quantifiable and reproducible way from images acquired with perfluoropropane contrast, as well as investigating how these extracted measures compare to those extracted from sulphur hexafluoride contrast and non-contrast studies. Methods 100 patients who received perfluoropropane contrast during their stress echocardiogram were recruited. Their stress echocardiography images were processed through a deep learning algorithm. Novel feature values were recorded and a subset of 10 studies were repeated. The automated measures of global longitudinal strain (GLS) and ejection fraction (EF) extracted from these images were compared to values previously extracted from sulphur hexafluoride contrast and non-contrast images using the same software. Results A full set of 31 novel imaging features were successfully extracted from 79 studies acquired using the perfluoropropane contrast agent with a dropout rate of 14% (n=92, 8 incomplete image sets). Repeated analysis in a subset of 10 perfluoropropane cases demonstrated excellent reproducibility of the extracted feature values (R2=1). Automated values of GLS and EF, at both rest (GLS = −16.4±4.8%, EF = 63±13%) and stress stages (GLS = −17.7±5.8%, EF = 68±11%), were extracted from 83 perfluoropropane studies, with a dropout rate of 16% (n=99, fewer incomplete sets as short axis view not required). The ranges of GLS and EF measures extracted from the perfluoropropane images were comparable to the other contrast studies (n=222) (Rest GLS = −16.8±5.8%, Rest EF = 63±10%; Stress GLS = −19.1±6.7%, Stress EF = 71±9%) and non-contrast studies (n=86) (Rest GLS = −15.7±5.3%, Rest EF = 57±10%; Stress GLS = −17.3±6.4%, Stress EF = 61±14%). Conclusions Novel features and clinically relevant measures were extracted from images acquired using perfluoropropane contrast for the first time in a fully automated and reproducible way using a deep learning algorithm. The analysis failure rate and generated measures are comparable to those extracted from images using other commonly used sulphur hexafluoride contrast agents and non-contrast stress echocardiography studies. These findings demonstrate that deep learning algorithms can be used for automated quantitative analysis of stress echocardiograms acquired using various contrast agents and in non-contrast studies to improve stress echocardiography practice. Funding Acknowledgement Type of funding source: Private company. Main funding source(s): Lantheus Medical Imaging, Inc.

Stretching force constants as descriptors of energy and geometry of F···HF hydrogen bonds.

In this work applicability of proton donor group stretching vibration force constants ks and intermolecular stretching force constants kσ for evaluations of hydrogen bond strength and geometry are discussed. For a set of 30 complexes with F···HF hydrogen bonds in a wide range 0.5-48kcal/mol by means of quantum chemical calculations equilibrium geometries, complexation energies, vibrational frequencies and corresponding force constants were calculated (MP2/aug-cc-pVTZ). It is shown, that properties of a hydrogen bond are more strictly correlated with the values of force constants than with vibrational frequencies. Easy-to-use equations for estimations of hydrogen bond energy ∆E and geometry (rFH, rFF) based on ks and kσ values are proposed.

Urban morphology parameterization for assessing the effects of housing blocks layouts on air temperature in the Euro-Mediterranean context

Urban environment characteristics directly affect local microclimate and, in turn, changes in local microclimate affect the urban environmental conditions, often resulting in negative consequences on human health and comfort. The study of these interactions is largely supported by scientific literature, but there is a general lack concerning the urban forms influence on outdoor thermal comfort in Euro-Mediterranean cities.In this study, the topic is addressed by investigating performances and characteristics of typical Euro-Mediterranean housing blocks to highlight their effects on local microclimate. Its relevance is related to the need for supporting the assessment of microclimate conditions and outdoor thermal comfort, providing an effective tool that can steer future adaptation policies and design processes.In this perspective, a parametric study is performed by investigating different housing blocks through some morphometric descriptors (density, land-use, canyon geometry descriptors), commonly used in urban and environmental planning. Seven typologies (detached, attached, linear, U-shaped, block with courtyard, block with multiple courtyards, tower) are modelled with the Computational Fluid Dynamic (CFD) ENVI-met v4.4 model, considering three configurations of surface cover and wind direction. For each combination, air temperature at pedestrian level (TA,1.6m) and descriptors are compared to highlight the microclimate performances of urban forms.The results show a close relationship between density, geometry and TA,1.6m distribution, highlighting the perimeter type (which includes three block types: U-shaped, block with courtyard, block with multiple courtyards) as the optimal configuration. Moreover, in some cases, a reduced amount of vegetation can lead to a significant reduction in air temperature, with economic benefits in terms of cost reductions for implementation, management and maintenance.

Statistical analysis and comparative study of multi-scale 2D and 3D shape features for unbound granular geomaterials

Morphological analysis is a prerequisite step for studying the mechanical properties of unbound granular geomaterials through both laboratory experiments and numerical simulations. In this study, two typical geomaterials, e.g., ballast and cobble particles, were sampled for study. First, a systematic framework was established to acquire two-dimensional (2D) particle outline and three-dimensional (3D) morphology for statistical shape analysis. The photographic-based system was set up to acquire 2D images of particle projections from multiple surrounding directions, then the images obtained were used to reconstruct 3D particle morphology based on close-range photogrammetry. Second, in order to quantify particle shape at three different geometric scales, 6 well-acknowledged 2D and 3D geometry descriptors, i.e., aspect ratio, sphericity, convexity, roundness, regularity, and roughness were computed. Third, the probability distribution function (PDF) of shape descriptors were fitted with different functions. Finally, 2D and 3D shape descriptors were statistically compared and the relationships between them were revealed by analyzing their cumulative distribution functions (CDFs).

A Comparative Classification Analysis of Abdominal Aortic Aneurysms by Machine Learning Algorithms.

The objective of this work was to perform image-based classification of abdominal aortic aneurysms (AAA) based on their demographic, geometric, and biomechanical attributes. We retrospectively reviewed existing demographics and abdominal computed tomography angiography images of 100 asymptomatic and 50 symptomatic AAA patients who received an elective or emergent repair, respectively, within 1-6 months of their last follow up. An in-house script developed within the MATLAB computational platform was used to segment the clinical images, calculate 53 descriptors of AAA geometry, and generate volume meshes suitable for finite element analysis (FEA). Using a third party FEA solver, four biomechanical markers were calculated from the wall stress distributions. Eight machine learning algorithms (MLA) were used to develop classification models based on the discriminatory potential of the demographic, geometric, and biomechanical variables. The overall classification performance of the algorithms was assessed by the accuracy, area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and precision of their predictions. The generalized additive model (GAM) was found to have the highest accuracy (87%), AUC (89%), and sensitivity (78%), and the third highest specificity (92%), in classifying the individual AAA as either asymptomatic or symptomatic. The k-nearest neighbor classifier yielded the highest specificity (96%). GAM used seven markers (six geometric and one biomechanical) to develop the classifier. The maximum transverse dimension, the average wall thickness at the maximum diameter, and the spatially averaged wall stress were found to be the most influential markers in the classification analysis. A second classification analysis revealed that using maximum diameter alone results in a lower accuracy (79%) than using GAM with seven geometric and biomechanical markers. We infer from these results that biomechanical and geometric measures by themselves are not sufficient to discriminate adequately between population samples of asymptomatic and symptomatic AAA, whereas MLA offer a statistical approach to stratification of rupture risk by combining demographic, geometric, and biomechanical attributes of patient-specific AAA.

The Evolution of Color Pattern in Butterflyfishes (Chaetodontidae).

Coral reef fishes constitute one of the most diverse assemblages of vertebrates on the planet. Color patterns are known to serve a number of functions including intra- and inter-specific signaling, camouflage, mimicry, and defense. However, the relative importance of these and other factors in shaping color pattern evolution is poorly understood. Here we conduct a comparative phylogenetic analysis of color pattern evolution in the butterflyfishes (Chaetodontidae). Using recently developed tools for quantifying color pattern geometry as well as machine learning approaches, we investigate the tempo of evolution of color pattern elements and test whether ecological variables relating to defense, depth, and social behavior predict color pattern evolution. Butterflyfishes exhibit high diversity in measures of chromatic conspicuousness and the degrees of fine versus gross scale color patterning. Surprisingly, most diversity in color pattern was not predicted by any of the measures of ecology in our study, although we did find a significant but weak relationship between the level of fine scale patterning and some aspects of defensive morphology. We find that the tempo of color pattern diversification in butterflyfishes has increased toward the present and suggest that rapid evolution, presumably in response to evolutionary pressures surrounding speciation and lineage divergence, has effectively decoupled color pattern geometry from some aspects of ecology. Machine learning classification of color pattern appears to rely on a set of features that are weakly correlated with current color pattern geometry descriptors, but that may be better suited for the detection of discrete components of color pattern. A key challenge for future studies lies in determining whether rapid evolution has generally decoupled color patterns from ecology, or whether convergence in function produces convergence in color pattern at phylogenetic scales.

3D geometry-based face recognition in presence of eye and mouth occlusions

This study proposes a novel occlusions detection and restoration strategy. The aim is to success with 3D face recognition even when faces are partially occluded by external objects. The method, which relies on geometrical facial properties, is designed for managing two types of facial occlusions (eye and mouth occlusions due to hands). First occlusions are detected and (if present) classified, by considering their effects on the 3D points cloud. Then, the occluded regions are progressively removed, and finally, the non-occluded symmetrical regions are used to restore the missing information. After the restoration process, face recognition is performed relying on the restored facial information and on the localized landmarks. The landmarking methodology relies on derivatives and on 12 differential geometry descriptors. The discriminating features adopted for facial comparison include shape index histograms, Euclidean and geodetical distances between landmarks, facial curves, and nose volume. Obtained recognition rates, evaluated on the whole Bosphorus database and on our private dataset, ranging from 92.55 to 97.20% depending on the completeness of data.

Classifying and analysis of random composites using structural sums feature vector.

The main goal of this paper is to present the application of structural sums, mathematical objects originating from the computational materials science, in construction of a feature space vector of two-dimensional random composites simulated by distributions of non-overlapping discs on the plane. Construction of the feature vector enables the immediate application of machine learning tools and data analysis techniques to random structures. In order to present the accuracy and the potential of structural sums as geometry descriptors, we apply them to classification problems comprising composites with circular inclusions as well as composites with shapes formed by discs. As an application, we perform the analysis of different models of composites in order to formulate the irregularity measure of random structures. We also visualize the relationship between the effective conductivity of two-dimensional composites and the geometry of inclusions.

Simple Coordination Geometry Descriptors Allow to Accurately Predict Metal-Binding Sites in Proteins.

With more than a third of the genome encoding for metal-containing biomolecules, the in silico prediction of how metal ions bind to proteins is crucial in chemistry, biology, and medicine. To date, algorithms for metal-binding site prediction are mainly based on sequence analysis. Those methods have reached enough quality to predict the correct region of the protein and the coordinating residues involved in metal-binding, but they do not provide three-dimensional (3D) models. On the contrary, the prediction of accurate 3D models for protein–metal adducts by structural bioinformatics and molecular modeling techniques is still a challenge. Here, we present an update of our multipurpose molecular modeling suite, GaudiMM, to locate metal-binding sites in proteins. The approach is benchmarked on 105 X-ray structures with resolution lower than 2.0 Å. Results predict the correct binding site of the metal in the biological scaffold for all the entries in the data set. Generated 3D models of the protein–metal coordination complexes reach root-mean-square deviation values under 1.0 Å between calculated and experimental structures. The whole process is purely based on finding poses that satisfy metal-derived geometrical rules without needing sequence or fine electronic inputs. Additional post-optimizations, including receptor flexibility, have been tested and suggest that more extensive searches, required when the host structures present a low level of pre-organization, are also possible. With this new update, GaudiMM is now able to look for metal-binding sites in biological scaffolds and clearly shows how explicitly considering the geometric particularities of the first coordination sphere of the metal in a docking process provides excellent results.

Application of geometry to RGB images for facial landmark localisation - a preliminary approach

This study proposes a novel approach to automatically localise 11 landmarks from facial RGB images. The novelty of this method relies on the application, i.e., point-by-point mapping, of 11 differential geometry descriptors such as curvatures to the three individual RGB image components. Thus, three-dimensional features are applied to bidimensional facial image representations and used, via thresholding techniques, to extract the landmark positions. The method was tested on the Bosphorus database and showed global average errors lower than five millimetres. The idea behind this approach is to embed this methodology in state-of-the-art 3D landmark detection methods to accomplish a full automatic landmarking by exploiting the advantages of both 2D and 3D data. Some landmarks such as pupils are arduous to be automatically extracted only via three-dimensional techniques. Thus, this method is intended as a bridging-the-gap preliminary technique that takes advantages of 2D imaging only for integrating advanced landmark localisation methodologies.

Application of geometry to RGB images for facial landmark localisation - a preliminary approach

This study proposes a novel approach to automatically localise 11 landmarks from facial RGB images. The novelty of this method relies on the application, i.e., point-by-point mapping, of 11 differential geometry descriptors such as curvatures to the three individual RGB image components. Thus, three-dimensional features are applied to bidimensional facial image representations and used, via thresholding techniques, to extract the landmark positions. The method was tested on the Bosphorus database and showed global average errors lower than five millimetres. The idea behind this approach is to embed this methodology in state-of-the-art 3D landmark detection methods to accomplish a full automatic landmarking by exploiting the advantages of both 2D and 3D data. Some landmarks such as pupils are arduous to be automatically extracted only via three-dimensional techniques. Thus, this method is intended as a bridging-the-gap preliminary technique that takes advantages of 2D imaging only for integrating advanced landmark localisation methodologies.

Habitat geometry, a step toward general bird community assembly rules in mature forests

Community ecology has often been criticized because it seldom defines general rules that can be exported outside of the studied system. Although much progress has been achieved by studying functional traits instead of species identities per se, environmental gradients defined in terms of local resources can be measured only on a subset of ecosystems and, thus, are not deemed general. Here, we used state-of-the-art statistical approaches for modeling the interaction between species functional traits (body size, feeding substrate) and habitat geometry (object size, shape and color; packing, layering and texturing), such as to derive general assembly rules for bird communities in an even-aged mature forest (La Mauricie National Park, Québec, Canada). Our results show that habitat geometry filters the species pool through its interaction with bird body size and feeding substrate, both at the species, and even more so at the community level. We found comparable assembly rules using only photographic (image-based) descriptors of habitat geometry, thus providing forest managers with a fast, reliable, standardized and cost-effective protocol for characterizing forest stands. Because bird functional groups respond differently to environmental gradients, we conclude that mature forest should be managed to preserve a spatial mosaic of successional states.

Numerical modelling of moisture diffusion in FRP with clustered microstructures

This paper presents a numerical study of moisture propagation in fibre reinforced polymer (FRP) materials with spatial tortuosity. An algorithm for creation of microstructures with clustered fibre architecture is presented. Several controls for different characteristics of the clusters are described. Different statistical descriptors of geometry are used to quantify the clustering. A Fickian diffusion process has been modelled. The effect of clustering on moisture diffusion through the FRP is reported. It is observed that microstructures greatly affect the diffusion behaviour of the FRP. Correlation between moisture diffusion and the statistical descriptors has been examined. The statistical measures that depict the global characteristics of the microstructure are found to have a better correlation with time to saturate the representative volume element. The results should help in prediction of macroscopic response and durability of FRPs.

Intramolecular O-H···O═C hydrogen bond energy via the molecular tailoring approach to RAHB structures.

A method for the calculation of the intramolecular hydrogen bond (HB) energy (EHB) by molecular tailoring approach for hydroxycarbonyl aliphatic compounds has been used for compounds with resonance-assisted hydrogen bonding (RAHB). The intramolecular hydrogen bond energies estimated for 229 structures (of 186 compounds) range from 8.2 to 26.3 kcal/mol and show correlation with the geometry descriptors of hydrogen bonds, with the calculated frequencies as well as with topological parameters obtained from the atoms in molecules (AIM) theory. These correlations differ significantly from obtained formerly for saturated nonenolizable structures and prove the special character of the resonance-assisted hydrogen-bonded systems.

Estimation of the intramolecular O-H···O═C hydrogen bond energy via the molecular tailoring approach. Part I: aliphatic structures.

A simple and universal method for the estimation of the intramolecular hydrogen bond (HB) energy (E(HB)) in hydroxycarbonyl aliphatic compounds is proposed by the application of the molecular tailoring approach (MTA) based on calculations at the second-order Møller-Plesset MP2 level. The calculation of EHB can be realized by the one optimization and three single point calculations of the energy for each compound with carbonyl and hydroxyl groups involved in HB. The intramolecular hydrogen bond energies estimated for 153 structures (of 102 compounds) ranged from 1.4 to 13.7 kcal/mol for systems without resonance-assisted hydrogen bonding (RAHB). To verify the method, we show the correlations of the energy (E(HB)) in six-, seven-, and eight-membered HB rings in the optimized multifunctional molecules with the usual geometry descriptors of hydrogen bonds. Moreover, topological parameters from the atoms in molecules (AIM) theory and the calculated infrared and proton NMR spectra are correlated. The effects of conjugation and π-electron delocalization, bifurcation, and cooperativity are discussed, along with the correlation between the strength and geometrical parameters of H bonding.

Integration of intensity textures and local geometry descriptors from Terrestrial Laser Scanning to map chert in outcrops

The potential of Terrestrial Laser Scanner imaging (TLS) as a tool to map chert, an amorphous variety of silica diffused in sedimentary rocks, is here discussed together with an original method for its automatic detection. Reflectance measurements in the VIS-NIR band (400–2500nm) show that chert displays low reflectance in the IR wavelengths that are operated by several commercial TLS. To develop and test a recognition method an outcrop of limestone with chert nodules was scanned with an IR (1541nm) TLS. The intensity information, after proper distance correction, was coupled with geometric and intensity descriptors for training Support Vector Machines (SVM) to separate vegetation from rock and limestone from chert. Results, cross inspected in the field and with reference pictures, demonstrate that TLS data can be efficiently exploited to map chert when the monochromatic information of the intensity is integrated with feature descriptors and SVM classifiers.