MolNet_Equi: A Chemically Intuitive, Rotation-Equivariant Graph Neural Network.

Although deep-learning (DL) models suggest unprecedented prediction capabilities in tackling various chemical problems, their demonstrated tasks have so far been limited to the scalar properties including the magnitude of vectorial properties, such as molecular dipole moments. A rotation-equivariant MolNet_Equi model, proposed in this paper, understands and recognizes the molecular rotation in the 3D Euclidean space, and exhibits the ability to predict directional dipole moments in the rotation-sensitive mode, as well as showing superior performance for the prediction of scalar properties. Three consecutive operations of molecular rotation R M ${\left(R\left(M\right)\right)}$ , dipole-moment prediction φ µ R M ${\left({\phi{} }_{\mu }\left(R\left(M\right)\right)\right)}$ , and dipole-moment inverse-rotation R - 1 φ µ R M ${\left({R}^{-1}\left({\phi{} }_{\mu }\left(R\left(M\right)\right)\right)\right)}$ do not alter the original prediction of the total dipole moment of a molecule φ µ M ${\left({\phi{} }_{\mu }\right(M\left)\right)}$ , assuring the rotational equivariance of MolNet_Equi. Furthermore, MolNet_Equi faithfully predicts the absolute direction of dipole moments given molecular poses, albeit the model has been trained only with the information on dipole-moment magnitudes, not directions. This work highlights the potential of incorporating fundamental yet crucial chemical rules and concepts into DL models, leading to the development of chemically intuitive models.

In vitro induction of in vivo-relevant stellate astrocytes in 3D brain-derived, decellularized extracellular matrices.

In vitro fabrication of 3D cell culture systems that could provide in vivo tissue-like, structural, and biochemical environments to neural cells is essential not only for fundamental studies on brain function and behavior, but also for tissue engineering and regenerative medicine applicable to neural injury and neurodegenerative diseases. In particular, for astrocytes-which actively respond to the surroundings and exhibit varied morphologies based on stimuli (e.g., stiffness and chemicals) in vitro, as well as physiological or pathological conditions in vivo-it is crucial to establish an appropriate milieu in in vitro culture platforms. Herein, we report the induction of in vivo-relevant, stellate-shaped astrocytes derived from cortices of Rattus norvegicus by constructing the 3D cell culture systems of brain-derived, decellularized extracellular matrices (bdECMs). The bdECM hydrogels were mechanically stable and soft, and the bdECM-based 3D scaffolds supplied biochemically active environments that astrocytes could interact with, leading to the development of in vivo-like stellate structures. In addition to the distinct morphology with actively elongated endfeet, the astrocytes, cultured in 3D bdECM scaffolds, would have neurosupportive characteristics, indicated by the accelerated neurite outgrowth in the astrocyte-conditioned media. Furthermore, next-generation sequencing showed that the gene expression profiles of astrocytes cultured in bdECMs were significantly different from those cultured on 2D surfaces. The stellate-shaped astrocytes in the bdECMs were analyzed to have reached a more mature state, for instance, with decreased expression of genes for scaffold ECMs, actin filaments, and cell division. The results suggest that the bdECM-based 3D culture system offers an advanced platform for culturing primary cortical astrocytes and their mixtures with other neural cells, providing a brain-like, structural and biochemical milieu that promotes the maturity and in vivo-like characteristics of astrocytes in both form and gene expression. STATEMENT OF SIGNIFICANCE: Decellularized extracellular matrices (dECMs) have emerged as strong candidates for the construction of three-dimensional (3D) cell cultures in vitro, owing to the potential to provide native biochemical and physical environments. In this study, we fabricated hydrogels of brain-derived dECMs (bdECMs) and cultured primary astrocytes within the bdECM hydrogels in a 3D context. The cultured astrocytes exhibited a stellate morphology distinct from conventional 2D cultures, featuring tridimensionally elongated endfeet. qRT-PCR and NGS-based transcriptomic analyses revealed gene expression patterns indicative of a more mature state, compared with the 2D culture. Moreover, astrocytes cultured in bdECMs showed neurosupportive characteristics, as demonstrated by the accelerated neurite outgrowth in astrocyte-conditioned media. We believe that the bdECM hydrogel-based culture system can serve as an in vitro model system for astrocytes and their coculture with other neural cells, holding significant potential for neural engineering and therapeutic applications.

Turbulence strength C n2 estimation from video using physics-based deep learning.

Images captured from a long distance suffer from dynamic image distortion due to turbulent flow of air cells with random temperatures, and thus refractive indices. This phenomenon, known as image dancing, is commonly characterized by its refractive-index structure constant C n2 as a measure of the turbulence strength. For many applications such as atmospheric forecast model, long-range/astronomy imaging, and aviation safety, optical communication technology, C n2 estimation is critical for accurately sensing the turbulent environment. Previous methods for C n2 estimation include estimation from meteorological data (temperature, relative humidity, wind shear, etc.) for single-point measurements, two-ended pathlength measurements from optical scintillometer for path-averaged C n2, and more recently estimating C n2 from passive video cameras for low cost and hardware complexity. In this paper, we present a comparative analysis of classical image gradient methods for C n2 estimation and modern deep learning-based methods leveraging convolutional neural networks. To enable this, we collect a dataset of video capture along with reference scintillometer measurements for ground truth, and we release this unique dataset to the scientific community. We observe that deep learning methods can achieve higher accuracy when trained on similar data, but suffer from generalization errors to other, unseen imagery as compared to classical methods. To overcome this trade-off, we present a novel physics-based network architecture that combines learned convolutional layers with a differentiable image gradient method that maintains high accuracy while being generalizable across image datasets.

MolNet: A Chemically Intuitive Graph Neural Network for Prediction of Molecular Properties.

Most graph neural networks (GNNs) in deep-learning chemistry collect and update atom and molecule features from the fed atom (and, in some cases, bond) features, basically based on the two-dimensional (2D) graph representation of 3D molecules. However, the 2D-based models do not faithfully represent 3D molecules and their physicochemical properties, exemplified by the overlooked field effect that is a "through-space" effect, not a "through-bond" effect. We propose a GNN model, denoted as MolNet, which accommodates the 3D non-bond information in a molecule, via a noncovalent adjacency matrix A ‾ , and also bond-strength information from a weighted bond matrix B . Comparative studies show that MolNet outperforms various baseline GNN models and gives a state-of-the-art performance in the classification task of BACE dataset and regression task of ESOL dataset. This work suggests a future direction for the construction of deep-learning models that are chemically intuitive and compatible with the existing chemistry concepts and tools.

Rotational Variance-Based Data Augmentation in 3D Graph Convolutional Network.

This work proposes the data augmentation by molecular rotation, with consideration that the protein-ligand binding events are rotation-variant. As a proof-of-concept, known active (i. e., 1-labeled) ligands to human ß-secretase 1 (BACE-1) are rotated for the generation of 0-labeled data, and the rotation-dependent prediction accuracy of 3D graph convolutional network (3DGCN) is investigated after data augmentation. The data augmentation makes the orientation-recognizing ability of 3DGCN improved significantly in the classification task for BACE-1/ligand binding. Furthermore, the data-augmented 3DGCN has a capability for predicting active ligands from a candidate dataset, via improved performance of orientation recognition, which would be applied to virtual drug screening and discovery.

Precision Mapping of COVID-19 Vulnerable Locales by Epidemiological and Socioeconomic Risk Factors, Developed Using South Korean Data.

COVID-19 has severely impacted socioeconomically disadvantaged populations. To support pandemic control strategies, geographically weighted negative binomial regression (GWNBR) mapped COVID-19 risk related to epidemiological and socioeconomic risk factors using South Korean incidence data (January 20, 2020 to July 1, 2020). We constructed COVID-19-specific socioeconomic and epidemiological themes using established social theoretical frameworks and created composite indexes through principal component analysis. The risk of COVID-19 increased with higher area morbidity, risky health behaviours, crowding, and population mobility, and with lower social distancing, healthcare access, and education. Falling COVID-19 risks and spatial shifts over three consecutive time periods reflected effective public health interventions. This study provides a globally replicable methodological framework and precision mapping for COVID-19 and future pandemics.

Diagnostic serial interval as a novel indicator for contact tracing effectiveness exemplified with the SARS-CoV-2/COVID-19 outbreak in South Korea.

BACKGROUND: The clinical onset serial interval is often used as a proxy for the transmission interval of an infectious disease. For SARS-CoV-2/COVID-19, data on clinical onset serial intervals is limited, since symptom onset dates are not routinely recorded and do not exist in asymptomatic carriers. METHODS: We define the diagnostic serial interval as the time between the diagnosis dates of the infector and infectee. Based on the DS4C project data on SARS-CoV-2/COVID-19 in South Korea, we estimate the means of the diagnostic serial interval, the clinical onset serial interval, and the difference between the two. We use the balanced cluster bootstrap method to construct 95% bootstrap confidence intervals. RESULTS: The mean of the diagnostic serial interval was estimated to be 3.63 days (95% CI: 3.24, 4.01). The diagnostic serial interval was significantly shorter than the clinical onset serial interval (estimated mean difference -1.12 days, 95% CI: -1.98, -0.26). CONCLUSIONS: The relatively short diagnostic serial intervals of SARS-CoV-2/COVID-19 in South Korea are likely due to the country's extensive efforts towards contact tracing. We propose the mean diagnostic serial interval as a new indicator for the effectiveness of a country's contact tracing as part of the epidemic surveillance.