Accelerated Design for High-Entropy Alloys Based on Machine Learning and Multiobjective Optimization.

High-entropy alloys (HEAs) with high hardness and high ductility can be considered as candidates for wear-resistant applications. However, designing novel HEAs with multiple desired properties using traditional alloy design methods remains challenging due to the enormous composition space. In this work, we proposed a machine-learning-based framework to design HEAs with high Vickers hardness (H) and high compressive fracture strain (D). Initially, we constructed data sets containing 172,467 data with 161 features for D and H, respectively. Four-step feature selection was performed, with the selection of 12 and 8 features for the D and H prediction models based on the optimal algorithms of the support vector machine (SVR) and light gradient boosting machine (LightGBM), respectively. The R2 of the well-trained models reached 0.76 and 0.90 for the 10-fold cross validation. Nondominated sorting genetic algorithm version II (NSGA-II) and virtual screening were employed to search for the optimal alloying compositions, and four recommended candidates were synthesized to validate our methods. Notably, the D of three candidates have shown significant improvements compared to the samples with similar H in the original data sets, with increases of 135.8, 282.4, and 194.1% respectively. Analyzing the candidates, we have recommended suitable atomic percentage ranges for elements such as Al (2-14.8 at %), Nb (4-25 at %), and Mo (3-9.9 at %) in order to design HEAs with high hardness and ductility.

Disaggregated impacts of off-farm work participation on household vulnerability to food poverty in Ghana.

This study examines disaggregated impacts of participation in off-farm employment on household vulnerability to food poverty in Ghana. We use household-level data collected from smallholder farmers in Ghana. This study employs the multinomial endogenous switching regression model to account for selection bias due to both observed and unobserved heterogeneity. Our results indicate that participation in off-farm employment activities, such as petty trading, significantly decreases household vulnerability to food poverty. Our findings further show that households that do participate in arts and crafts as an off-farm activity are more vulnerable to food poverty had they not participated. This paper provides useful policy insights to enable smallholders involved in off-farm work activities to improve food consumption expenditure and reduce their risk of food poverty.

Search for ABO3 Type Ferroelectric Perovskites with Targeted Multi-Properties by Machine Learning Strategies.

Ferroelectric perovskites are one of the most promising functional materials due to the pyroelectric and piezoelectric effect. In the practical applications of ferroelectric perovskites, it is often necessary to meet the requirements of multiple properties. In this work, a multiproperties machine learning strategy was proposed to accelerate the discovery and design of new ferroelectric ABO3-type perovskites. First, a classification model was constructed with data collected from publications to distinguish ferroelectric and nonferroelectric perovskites. The classification accuracies of LOOCV and the test set are 87.29% and 86.21%, respectively. Then, two machine learning strategies, Machine-Learning Workflow and SISSO, were used to construct the regression models to predict the specific surface area (SSA), band gap (Eg), Curie temperature (Tc), and dielectric loss (tan δ) of ABO3-type perovskites. The correlation coefficients of LOOCV in the optimal models for SSA, Eg, and Tc are 0.935, 0.891, and 0.971, respectively, while the correlation coefficient of the predicted and experimental values of the SISSO model for tan δ prediction could reach 0.913. On the basis of the models, 20 ABO3 ferroelectric perovskites with three different application prospects were screened out with the required properties, which could be explained by the patterns between the important descriptors and the properties by using SHAP. Furthermore, the constructed models were developed into web servers for the researchers to accelerate the rational design and discovery of ABO3 ferroelectric perovskites with desired multiple properties.

Inferring novel genes related to colorectal cancer via random walk with restart algorithm.

Colorectal cancer (CRC) is the third most common type of cancer. In recent decades, genomic analysis has played an increasingly important role in understanding the molecular mechanisms of CRC. However, its pathogenesis has not been fully uncovered. Identification of genes related to CRC as complete as possible is an important way to investigate its pathogenesis. Therefore, we proposed a new computational method for the identification of novel CRC-associated genes. The proposed method is based on existing proven CRC-associated genes, human protein-protein interaction networks, and random walk with restart algorithm. The utility of the method is indicated by comparing it to the methods based on Guilt-by-association or shortest path algorithm. Using the proposed method, we successfully identified 298 novel CRC-associated genes. Previous studies have validated the involvement of the majority of these 298 novel genes in CRC-associated biological processes, thus suggesting the efficacy and accuracy of our method.

Machine-learning-assisted prediction of surgical outcomes in patients undergoing gastrectomy.

OBJECTIVE: Postoperative complications adversely affected the prognosis in patients with gastric cancer. This study intends to investigate the feasibility of using machine-learning model to predict surgical outcomes in patients undergoing gastrectomy. METHODS: In this study, cancer patients who underwent gastrectomy at Shanghai Rui Jin Hospital in 2017 were randomly assigned to a development or validation cohort in a 9:1 ratio. A support vector classification (SVC) model to predict surgical outcomes in patients undergoing gastrectomy was developed and further validated. RESULTS: A total of 321 patients with 32 features were collected. The positive and negative outcomes of postoperative complication after gastrectomy appeared in 100 (31.2%) and 221 (68.8%) patients, respectively. The SVC model was constructed to predict surgical outcomes in patients undergoing gastrectomy. The accuracy of 10-fold cross validation and external verification was 78.17% and 78.12%, respectively. Further, an online web server has been developed to share the SVC model for machine-learning-assisted prediction of surgical outcomes in patients undergoing gastrectomy in the future procedures, which is accessible at the web address: http://47.100.47.97:5005/r_model_prediction. CONCLUSIONS: The SVC model was a useful predictor for measuring the risk of postoperative complications after gastrectomy, which may help stratify patients with different overall status for choice of surgical procedure or other treatments. It can be expected that machine-learning models in cancer informatics research are possibly shareable and accessible via web address all over the world.

Prediction of potential drivers connecting different dysfunctional levels in lung adenocarcinoma via a protein-protein interaction network.

Lung cancer is a serious disease that threatens an affected individual's life. Its pathogenesis has not yet to be fully described, thereby impeding the development of effective treatments and preventive measures. "Cancer driver" theory considers that tumor initiation can be associated with a number of specific mutations in genes called cancer driver genes. Four omics levels, namely, (1) methylation, (2) microRNA, (3) mutation, and (4) mRNA levels, are utilized to cluster cancer driver genes. In this study, the known dysfunctional genes of these four levels were used to identify novel driver genes of lung adenocarcinoma, a subtype of lung cancer. These genes could contribute to the initiation and progression of lung adenocarcinoma in at least two levels. First, random walk with restart algorithm was performed on a protein-protein interaction (PPI) network constructed with PPI information in STRING by using known dysfunctional genes as seed nodes for each level, thereby yielding four groups of possible genes. Second, these genes were further evaluated in a test strategy to exclude false positives and select the most important ones. Finally, after conducting an intersection operation in any two groups of genes, we obtained several inferred driver genes that contributed to the initiation of lung adenocarcinoma in at least two omics levels. Several genes from these groups could be confirmed according to recently published studies. The inferred genes reported in this study were also different from those described in a previous study, suggesting that they can be used as essential supplementary data for investigations on the initiation of lung adenocarcinoma. This article is part of a Special Issue entitled: Accelerating Precision Medicine through Genetic and Genomic Big Data Analysis edited by Yudong Cai & Tao Huang.

Using Data Mining To Search for Perovskite Materials with Higher Specific Surface Area.

The specific surface area (SSA) of ABO3-type perovskite is one of the important properties associated with photocatalytic ability. In this work, data mining methods were used to explore the relationship between the SSA (in the range of 1-60 m2 g-1) of perovskite and its features, including chemical compositions and technical parameters. The genetic algorithm-support vector regression method was used to screen the main features for modeling. The correlation coefficient ( R) between the predicted and experimental SSAs reached as high as 0.986 for the training data set and 0.935 for leave-one-out cross-validation. ABO3-type perovskites with higher SSA can be screened out using the Online Computation Platform for Materials Data Mining (OCPMDM) developed in our laboratory. Further, an online web server has been developed to share the model for the prediction of SSA of ABO3-type perovskite, which is accessible at http://118.25.4.79/material_api/csk856q0fulhhhwv .

Predicting the toxicities of metal oxide nanoparticles based on support vector regression with a residual bootstrapping method.

For safely using the untested metal oxide nanoparticles (MONPs) in industrial and commercial applications, it is important to predict their potential toxicities quickly and efficiently. In this research, the quantitative structure-activity relationship (QSAR) model based on support vector regression (SVR) with a residual bootstrapping technique (BTSVR) was proposed to predict the toxicities of MONPs. It was found that the main features influencing the toxicities of MONPs were RAatom (atomic ratio of oxygen to metal), ΔHm (enthalpy of melting), and Ecoh (cohesive energy). The QSPR model constructed was robust and self-explanatory in predicting the toxicities of MONPs with the coefficient of determination (R2) of 0.87 and the root mean square error (RMSE) of 0.184 for the training sets, and R2 of 0.84 and RMSE of 0.217 for the testing sets, respectively. The performance of our model is much better than that published. Moreover, our model was validated by the external testing sets 1000 times. Therefore, it is expected that the method presented here can be used to construct powerful model in predicting the toxicities of MONPs untested or even unavailable.

Mining for genes related to choroidal neovascularization based on the shortest path algorithm and protein interaction information.

BACKGROUND: Choroidal neovascularization (CNV) is a serious eye disease that may cause visual loss, especially for older people. Many factors have been proven to induce this disease including age, gender, obesity, and so on. However, until now, we have had limited knowledge on CNV's pathogenic mechanism. Discovering the genes that underlie this disease and performing extensive studies on them can help us to understand how CNV occurs and design effective treatments. METHODS: In this study, we designed a computational method to identify novel CNV-related genes in a large protein network constructed using the protein-protein interaction information in STRING. The candidate genes were first extracted from the shortest paths connecting any two known CNV-related genes and then filtered by a permutation test and using knowledge of their linkages to known CNV-related genes. RESULTS: A list of putative CNV-related candidate genes was accessed by our method. These genes are deemed to have strong relationships with CNV. CONCLUSIONS: Extensive analyses of several of the putative genes such as ANK1, ITGA4, CD44 and others indicate that they are related to specific biological processes involved in CNV, implying they may be novel CNV-related genes. GENERAL SIGNIFICANCE: The newfound putative CNV-related genes may provide new insights into CNV and help design more effective treatments. This article is part of a Special Issue entitled "System Genetics" Guest Editor: Dr. Yudong Cai and Dr. Tao Huang.

Electrolyte-controlled discharge product distribution of Na-O2 batteries: a combined computational and experimental study.

Tuning the composition of discharge products is an important strategy to reduce charge potential, suppress side reactions, and improve the reversibility of metal-oxygen batteries. In the present study, first-principles calculations and experimental confirmation were performed to unravel the influence of O2 pressure, particle size, and electrolyte on the composition of charge products in Na-O2 batteries. The electrolytes with medium and high donor numbers (>12.5) are favorable for the formation of sole NaO2, while those with low donor numbers (<12.5) may permit the formation of Na2O2 by disproportionation reactions. Our comparative experiments under different electrolytes confirmed the calculation prediction. Our calculations indicated that O2 pressure and particle size hardly affect discharge products. On the electrode, only one-electron-transfer electrochemical reaction to form NaO2 takes place, whereas two-electron-transfer electrochemical and chemical reactions to form Na2O2 and Na3O4 are prevented in thermodynamics. The present study explains why metastable NaO2 was identified as a sole discharge product in many experiments, while thermodynamically more stable Na2O2 was not observed. Therefore, to achieve low overpotential, a high-donor-number electrolyte should be applied in the discharge processes of Na-O2 batteries.

Predicting network of drug-enzyme interaction based on machine learning method.

It is important to correctly and efficiently map drugs and enzymes to their possible interaction network in modern drug research. In this work, a novel approach was introduced to encode drug and enzyme molecules with physicochemical molecular descriptors and pseudo amino acid composition, respectively. Based on this encoding method, Random Forest was adopted to build the drug-enzyme interaction network. After selecting the optimal features that are able to represent the main factors of drug-enzyme interaction in our prediction, a total of 129 features were attained which can be clustered into nine categories: Elemental Analysis, Geometry, Chemistry, Amino Acid Composition, Secondary Structure, Polarity, Molecular Volume, Codon Diversity and Electrostatic Charge. It is further found that Geometry features were the most important of all the features. As a result, our predicting model achieved an MCC of 0.915 and a sensitivity of 87.9% at the specificity level of 99.8% for 10-fold cross-validation test, and achieved an MCC of 0.895 and a sensitivity of 95.7% at the specificity level of 95.4% for independent set test. This article is part of a Special Issue entitled: Computational Proteomics, Systems Biology & Clinical Implications. Guest Editor: Yudong Cai.

Theoretical study of acene-bridged dyes for dye-sensitized solar cells.

The electronic structures and absorption spectra for a series of acene-based organic dyes and the adsorption energy and optical properties for these dyes adsorbed on (TiO2)38 have been investigated using density functional theory (DFT) and time-dependent density functional theory (TDDFT) methods. The effects of acene units and different substitution positions of electron donors on the optoelectronic properties of the acene-modified dyes are demonstrated. The photophysical properties of tetracene- and pentacene-based dyes are found to be tuned by changing the size of acene and the substitution position of the donor. The donor sites have a significant influence on the absorption wavelength mainly because of different molecular orbital (MO) contributions of the highest occupied molecular orbital (HOMO) on the bridging acene units, and the increasing MO contribution would lead to the red shift in the absorption spectra. Meanwhile, the donor is located close to the center of the π-conjugated bridge, and the absorption spectra are extended. The adsorption energy and optical properties of tetracene- and pentacene-based dyes adsorbed on (TiO2)38 suggest that acene-bridged dyes could be adsorbed on the TiO2 surface and inject electrons into semiconductors effectively. Then the results obtained from the hexacene-based dyes confirm the conclusions proposed from the tetracene- and pentence-based dyes. This study will provide a useful reference to the future design and optimization of acene dyes for dye-sensitized solar cell applications.

Inverse Design of Low-Resistivity Ternary Gold Alloys via Interpretable Machine Learning and Proactive Search Progress.

Ternary gold alloys (TGAs) are highly regarded for their excellent electrical properties. Electrical resistivity is a crucial indicator for evaluating the electrical performance of TGAs. To explore new promising TGAs with lower resistivity, we developed a reverse design approach integrating machine learning techniques and proactive searching progress (PSP) method. Compared with other models, the support vector regression (SVR) was determined to be the most optimal model for resistivity prediction. The training and test sets yielded R2 values of 0.73 and 0.77, respectively. The model interpretation indicated that lower electrical resistivity was associated with the following conditions: a van der Waals Radius (Vrt) of 0, a Vr (another van der Waals Radius) of less than 217, and a mass attenuation coefficient of MoKα (Macm) greater than 77.5 cm2g-1. Applying the PSP method, we successfully identified eight candidates whose resistivity was lower than that of the sample with the lowest resistivity in the dataset by more than 53-60%, e.g., Au1.000Cu4.406Pt1.833 and Au1.000Pt2.232In1.502. Finally, the candidates were validated to possess low resistivity through the pattern recognition method.

Feature Selection in Machine Learning for Perovskite Materials Design and Discovery.

Perovskite materials have been one of the most important research objects in materials science due to their excellent photoelectric properties as well as correspondingly complex structures. Machine learning (ML) methods have been playing an important role in the design and discovery of perovskite materials, while feature selection as a dimensionality reduction method has occupied a crucial position in the ML workflow. In this review, we introduced the recent advances in the applications of feature selection in perovskite materials. First, the development tendency of publications about ML in perovskite materials was analyzed, and the ML workflow for materials was summarized. Then the commonly used feature selection methods were briefly introduced, and the applications of feature selection in inorganic perovskites, hybrid organic-inorganic perovskites (HOIPs), and double perovskites (DPs) were reviewed. Finally, we put forward some directions for the future development of feature selection in machine learning for perovskite material design.

Discovery of ANO1 Inhibitors based on Machine learning and molecule docking simulation approaches.

Calcium-activated chloride channels (CaCCs) are chloride channels that are regulated according to intracellular calcium ion concentrations. The channel protein ANO1 is widely present in cells and is involved in physiological activities including cellular secretion, signaling, cell proliferation and vasoconstriction and diastole. In this study, the ANO1 inhibitors were investigated with machine learning and molecular simulation. Two-dimensional structure-activity relationship (2D-SAR) and three-dimensional quantitative structure-activity relationship (3D-QSAR) models were developed for the qualitative and quantitative prediction of ANO1 inhibitors. The results showed that the prediction accuracies of the model were 85.9% and 87.8% for the training and test sets, respectively, and 85.9% and 87.8% for the rotating forest (RF) in the 2D-SAR model. The CoMFA and CoMSIA methods were then used for 3D QSAR modeling of ANO1 inhibitors, respectively. The q2 coefficients for model cross-validation were all greater than 0.5, implying that we were able to obtain a stable model for drug activity prediction. Molecular docking was further used to simulate the interactions between the five most promising compounds predicted by the model and the ANO1 protein. The total score for the docking results between all five compounds and the target protein was greater than 6, indicating that they interacted strongly in the form of hydrogen bonds. Finally, simulations of amino acid mutations around the docking cavity of the target proteins showed that each molecule had two or more sites of reduced affinity following a single mutation, indicating outstanding specificity of the screened drug molecules and their protein ligands.

Inverse Design of Hybrid Organic-Inorganic Perovskites with Suitable Bandgaps via Proactive Searching Progress.

Hybrid organic-inorganic perovskites (HOIPs) have shown the encouraging development in solar cells that have achieved excellent device performance. One of the most important issues has been focused on finding Pb-free candidates with suitable bandgaps, which could accelerate the commercialization of environmentally friendly HOIP-based cells. Herein, we propose a new inverse design method, proactive searching progress (PSP), to efficiently discover potential HOIPs from universal chemical space by combining machine learning (ML) techniques. Compared to the pioneering work on this topic, we carried out our ML study based on 1201 collected HOIP samples with experimental bandgaps rather than theoretical properties. On the basis of 25 selected features, a weighted voting regressor ML model was constructed to predict bandgaps of HOIPs. The model comprehensively embedded four submodels and performed the coefficient determinations of 0.95 for leaving-one-out cross-validation and 0.91 for testing set. The feature analysis revealed that the tolerance factor (t f) below 0.971 and the new tolerance factor (τf) in 3.75-4.09 contributed to lower bandgaps and vice versa. By applying the PSP method, the Pb-free HOIPs with optimal bandgaps were successfully designed from a generated chemical space comprising over 8.20 × 1018 combinations, which included 733848 candidates (e.g., Cs0.334FA0.266MA0.400Sn0.769Ge0.003Pd0.228Br0.164I2.836) with an optimal bandgap of 1.34 eV for single junction solar cells, 1511073 large-bandgap candidates (e.g., Cs0.392FA0.016MA0.592Cr0.383Sr0.347Sn0.270Br1.171I1.829) for top parts in tandem solar cells (TSCs), and 20242 low-bandgap ones (e.g., MA0.815FA0.185Sn0.927Ge0.073I3) for bottom cells in TSCs. Finally, three new HOIPs were synthesized with an average bandgap error 0.07 eV between predictions and experiments. We are convinced that the proposed PSP method and ML progress could facilitate the discovery of new promising HOIPs for photovoltaic devices with the desired properties.

Predicting Experimental Formability of Hybrid Organic-Inorganic Perovskites via Imbalanced Learning.

Hybrid organic-inorganic perovskites (HOIPs) have gained lots of attention in the photovoltaic field, but their further development is restrained by contaminant and stability. More potential HOIPs should be explored for photovoltaic devices. In this work, we collected 539 HOIPs and 24 non-HOIPs experimentally synthesized to explore novel compositions of HOIPs. An imbalanced learning was carried out, and the best classification model achieved a leaving-one-out cross-validation accuracy of 100.0% and a test accuracy of 96.1%. The A site atomic radii (ARA), A site ionic radius (IRA), and tolerance factor (tf) were identified as the most important features. ARA < 2.72 Å, IRA < 2.65 Å, and tf < 1.01 contributed to perovskite formability, and the formability possibilities of the corresponding samples were over 90.0%. Potential A site organic fragments were identified for perovskite solar cells, such as dimethylamine, hydroxylamine, hydrazine, etc. Finally, three new Sn-Ge mixed systems of HOIPs were successfully synthesized, which was consistent with the model predictions.

Machine Learning-Assisted Design of Yttria-Stabilized Zirconia Thermal Barrier Coatings with High Bonding Strength.

As a high-quality thermal barrier coating material, yttria-stabilized zirconia (YSZ) can effectively reduce the temperature of the collective materials to be used on the surface of gas turbine hot-end components. The bonding strength between YSZ and the substrate is also one of the most important factors for the applications. Herein, the Gaussian mixture model (GMM) and support vector regression (SVR) were used to construct a machine learning model between YSZ coating bonding strength and atmospheric plasma spraying (APS) process parameters. First, GMM was used to expand the original 8 data points to 400 with the R value of leave-one-out cross-validation improved from 0.690 to 0.990. Then, the specific effects of APS process parameters were explored through Shapley additive explanations and sensitivity analysis. Principal component analysis was used to explain the constructed model and obtain the optimized area with a high bonding strength. After experimental validation, the results showed that under the APS process parameters of a current of 617 A, a voltage of 65 V, a H2 flow of 3 L min-1, and a thickness of 200 µm, the bonding strength increased by more than 19% to 55.5 MPa compared with the original maximum value of 46.6 MPa, indicating that the constructed GMM-SVR model can accurately predict the bonding strength of YSZ coating.

Accelerated Discovery of Ternary Gold Alloy Materials with Low Resistivity via an Interpretable Machine Learning Strategy.

New ternary gold alloys with low resistivities (ρ) were screened out via an interpretable machine learning strategy by using the support vector regression (SVR) model integrated with SHAP analysis. The correlation coefficient (R) and the root mean square error (RMSE) of test set were 0.876 and 0.302, respectively, indicating the strong generalization ability of the model. The average ρ of top 10 candidates was 1.22×10-7 â€…Ω m, which was 41% lower than the known minimum of 2.08×10-7 â€…Ω m. The outputs of SVR model were analyzed with the critical SHAP values including first ionization energy of C-site (584 kJ ⋅ mol-1 ), electronegativity of C-site (1.72) and the second ionization energy of B-site (1135 kJ ⋅ mol-1 ), respectively. Moreover, an online web server was developed to share the model at http://materials-data-mining.com/onlineservers/wxdaualloy.

Fe-N-C single-atom nanozymes based sensor array for dual signal selective determination of antioxidants.

Machine learning algorithms as a powerful tool can efficiently utilize and process large quantities of data generated by high-throughput experiments in various fields. In this work, we used a general ionic salt-assisted synthesis method to prepare oxidase-like Fe-N-C SANs. The possible reason for the excellent enzyme-mimicking activity and affinity of Fe-N-C SANs was further verified by density functional theory calculations. Due to the remarkable oxidase-mimicking activity, the prepared Fe-N-C SANs were used to detect ascorbic acid (AA) with a detection limit of 0.5 µM. Based on the machine learning algorithms, we successfully distinguished six antioxidants (ascorbic acid, glutathione, L-cysteine, dithiothreitol, uric acid, and dopamine) with the same concentration by either one kind of Fe-N-C SANs or three kinds of different Fe-N-C SANs. The usefulness of the Fe-N-C SANs sensor arrays was further validated by the hierarchal cluster analysis, where they also can be correctly identified. More importantly, a SANs-based digital-image colorimetric sensor array has also been successfully constructed and thereby achieved visual and informative colorimetric analysis for practical samples out of the lab. This work not only provides a design synthesis method to prepare SANs but also combines machine learning algorithms with SANs sensors to identify analytes with similar properties, which can further expand to the detection of proteins and cells related to diseases in the future.