10-fold Cross-validation Results Research Articles

Prediction of Super-enhancers Based on Mean-shift Undersampling

Background: Super-enhancers are clusters of enhancers defined based on the binding occupancy of master transcription factors, chromatin regulators, or chromatin marks. It has been reported that super-enhancers are transcriptionally more active and cell-type-specific than regular enhancers. Therefore, it is necessary to identify super-enhancers from regular enhancers. A variety of computational methods have been proposed to identify super-enhancers as auxiliary tools. However, most methods use ChIP-seq data, and the lack of this part of the data will make the predictor unable to execute or fail to achieve satisfactory performance. Objective: The aim of this study is to propose a stacking computational model based on the fusion of multiple features to identify super-enhancers in both human and mouse species. Methods: This work adopted mean-shift to cluster majority class samples and selected four sets of balanced datasets for mouse and three sets of balanced datasets for human to train the stacking model. Five types of sequence information are used as input to the XGBoost classifier, and the average value of the probability outputs from each classifier is designed as the final classification result. Results: The results of 10-fold cross-validation and cross-cell-line validation prove that our method has superior performance compared to other existing methods. The source code and datasets are available at https://github.com/Cheng-Han-max/SE_voting. Conclusion: The analysis of feature importance indicates that Mismatch accounts for the highest proportion among the top 20 important features.

Hourly near-ground NO2 concentration retrieval from geostationary satellite observations

Abstract. Nitrogen dioxide (NO2) is an important contributor to the formation of acid rain, photochemical smog and aerosol particles, which seriously endangers public health. At present, remote sensing of polar-orbiting satellites is a conventional means to obtain large-scale NO2 distribution, but it cannot capture the rapid change because of long revisit periods. The Advanced Himawari Imager (AHI) on the Himawari-8 geostationary satellite has the advantage of high time resolution, which makes it possible to realize near-real-time atmospheric monitoring. Here, based on the absorption characteristics of NO2 in infrared radiation, hourly near-surface NO2 concentrations are retrieved based on the brightness temperature from AHI and auxiliary information such as meteorology and aerosol. The results of 10-fold cross-validation show that NO2 estimations from satellite are in good agreement with in-situ measurements, and their determination coefficient (R2) can reach 0.79. Due to different emission and atmospheric diffusion conditions at different time, the model performance presents a diurnal variation of high accuracy in the noon and afternoon but low accuracy in the morning. Based on the retrieval dataset, it is found that high NO2 concentrations are mainly concentrated in the densely populated and industrial areas such as the North China area. In addition, NO2 pollution mainly occurs in autumn and winter, and the average NO2 concentration in winter is about 1.63 times that in summer in 2021. This study provides a new insight for satellite retrieval of NO2, which is of great significance for real-time pollution monitoring and public health protection.

3D medical image segmentation based on semi-supervised learning using deep co-training

In recent years, artificial intelligence has been applied to 3D COVID-19 medical image diagnosis, which reduces detection costs and missed diagnosis rates with higher predictive accuracy, and diagnostic efficiency. However, the limited size and low quality of clinical 3D medical image samples have hindered the segmentation performance of 3D models. Therefore, we propose a 3D medical image segmentation model based on semi-supervised learning using co-training. Multi-view and multi-modal images are generated using spatial flipping and windowing techniques to enhance the spatial diversity of 3D image samples. A pseudo label generation module based on confidence-weights is employed to generate reliable pseudo labels for non-annotated data, thereby increasing the sample size and reducing overfitting. The proposed approach utilizes a three-stage training process: firstly, training a single network based on annotated data; secondly, incorporating non-annotated data to train a dual-modal network and generate pseudo labels; finally, jointly training six models in three dimensions using both annotated and pseudo labels generated from multi-view and multi-modal images, aiming to enhance segmentation accuracy and generalization performance. Additionally, a consistency regularization loss is applied to reduce noises and accelerate convergence of the training. Moreover, a heatmap visualization method is employed to focus on the attention of features at each stage of training, providing effective reference for clinical diagnosis. Experiments were conducted on an open dataset of 3D COVID-19 CT samples and a non-annotated dataset from TCIA, including 771 NIFTI-format CT images from 661 COVID-19 patients. The results of 5-fold cross-validation show that the proposed model achieves a segmentation accuracy of Dice=73.30 %, ASD=10.633, Sensitivity=63.00 %, and Specificity=99.60 %. Compared to various typical semi-supervised learning 3D segmentation models, it demonstrates better segmentation accuracy and generalization performance.

Research on Landsat 8 land surface temperature retrieval and spatial–temporal migration capabilities based on random forest model

Land surface temperature (LST) plays a crucial role in characterizing land surface processes and the energy balance. Accurate monitoring of spatial–temporal variations in LST holds great significance for global and regional climate change research. However, with the increasing of multi-source remote sensing data, it has become a challenging task to construct an LST model that reduces dependence on auxiliary data (such as land surface emissivity (LSE) and atmospheric water vapor content (WVC) synchronized with satellite observations. This study proposed an LST retrieval model based on random forest model (RFLSTR). The results of 10-fold cross-validation showed that the RFLSTR model had high modelling accuracy, the root mean square error (RMSE) and mean absolute error (MAE) were lower than 2 K. Landsat 8 images from the Beijing area and South Korea, were utilized to assess the spatial–temporal migration performance of the RFLSTR model. Additionally, Landsat 9 images from South Korea were employed to analyse the sensor migration performance of the RFLSTR model. In conclusion, the findings highlight that the proposed RFLSTR model achieve high retrieval accuracy without relying on auxiliary data (such as LSE and WVC) synchronized with satellite observations. Moreover, the RFLSTR model exhibits robust spatial–temporal capabilities, enhancing the efficiency of LST retrieval and facilitating the utilization of satellite data.

Identifying pathological groups from MRI in prostate cancer using graph representation learning

Multiparametric magnetic resonance imaging (mpMRI) plays a critical role in prostate cancer (PCa) diagnosis, aiding in clinical trial evaluation and personalized treatment planning. We propose a novel prediction approach based on graph representation learning (GRL) integrating mpMRI images for classifying International Society of Urological Pathology (ISUP) grade groups. We initially constructed a mpMRI image graph representation learning (MMIGRL) model using a graph-based representation fusion approach. After constructing a heterogeneous graph integrating T2-weighted and apparent diffusion coefficient (ADC) images, we applied an adaptive method that dynamically adjusted the edges within the graph to reveal relationships within the mpMRI images. Subsequently, we developed a residual graph convolution network (Res-GCN) model to establish the mapping relationship between the mpMRI fusion graph and ISUP grade groups. The results of 5-fold cross-validation demonstrated excellent overall performance, and the prediction results using independent testing datasets confirmed the model's generalizability. The results of ablation studies highlighted the positive impact of integrating T2-weighted and ADC images and the effectiveness of residual connections in learning hierarchical features within the graph structure. The results of comparative experiments validated the superiority of the proposed MMIGRL model compared with convolutional neural network (CNN) and vision transformer (ViT) models. This approach holds promise for accurate ISUP grade group classification in PCa diagnosis, contributing to improved patient care and clinical decision making.

IProm-Yeast: Prediction Tool for Yeast Promoters Based on ML Stacking

Background and Objective: Gene promoters play a crucial role in regulating gene transcription by serving as DNA regulatory elements near transcription start sites. Despite numerous approaches, including alignment signal and content-based methods for promoter prediction, accurately identifying promoters remains challenging due to the lack of explicit features in their sequences. Consequently, many machine learning and deep learning models for promoter identification have been presented, but the performance of these tools is not precise. Most recent investigations have concentrated on identifying sigma or plant promoters. While the accurate identification of Saccharomyces cerevisiae promoters remains an underexplored area. In this study, we introduced “iPromyeast”, a method for identifying yeast promoters. Using genome sequences from the eukaryotic yeast Saccharomyces cerevisiae, we investigate vector encoding and promoter classification. Additionally, we developed a more difficult negative set by employing promoter sequences rather than nonpromoter regions of the genome. The newly developed negative reconstruction approach improves classification and minimizes the amount of false positive predictions. Methods: To overcome the problems associated with promoter prediction, we investigate alternate vector encoding and feature extraction methodologies. Following that, these strategies are coupled with several machine learning algorithms and a 1-D convolutional neural network model. Our results show that the pseudo-dinucleotide composition is preferable for feature encoding and that the machine- learning stacking approach is excellent for accurate promoter categorization. Furthermore, we provide a negative reconstruction method that uses promoter sequences rather than non-promoter regions, resulting in higher classification performance and fewer false positive predictions. Results: Based on the results of 5-fold cross-validation, the proposed predictor, iProm-Yeast, has a good potential for detecting Saccharomyces cerevisiae promoters. The accuracy (Acc) was 86.27%, the sensitivity (Sn) was 82.29%, the specificity (Sp) was 89.47%, the Matthews correlation coefficient (MCC) was 0.72, and the area under the receiver operating characteristic curve (AUROC) was 0.98. We also performed a cross-species analysis to determine the generalizability of iProm-Yeast across other species. Conclusion: iProm-Yeast is a robust method for accurately identifying Saccharomyces cerevisiae promoters. With advanced vector encoding techniques and a negative reconstruction approach, it achieves improved classification accuracy and reduces false positive predictions. In addition, it offers researchers a reliable and precise webserver to study gene regulation in diverse organisms.

DF-Phos: Prediction of Protein Phosphorylation Sites by Deep Forest.

Phosphorylation is the most important and studied post-translational modification (PTM), which plays a crucial role in protein function studies and experimental design. Many significant studies have been performed to predict phosphorylation sites using various machine-learning methods. Recently, several studies have claimed that deep learning-based methods are the best way to predict the phosphorylation sites because deep learning as an advanced machine learning method can automatically detect complex representations of phosphorylation patterns from raw sequences and thus offers a powerful tool to improve phosphorylation site prediction. In this study, we report DF-Phos, a new phosphosite predictor based on the Deep Forest to predict phosphorylation sites. In DF-Phos, the feature vector taken from the CkSAApair method is as input for a Deep Forest framework for predicting phosphorylation sites. The results of 10-fold cross-validation show that the Deep Forest method has the highest performance among other available methods. We implemented a Python program of DF-Phos, which is freely available for non-commercial use at https://github.com/zahiriz/DF-Phos Moreover, users can use it for various PTM predictions.

Spatial prediction of lime requirements by adjusting aluminium saturation in Sub-Saharan Africa croplands

Acidic soils cover over 30 % of Sub-Saharan Africa cropland. Acidic soils deprive crops of calcium, magnesium, potassium, molybdenum, and phosphorus due to aluminium (Al), manganese, and iron toxicities. Thus, liming is required to adjust the level of exchangeable Al3+ to the desired level of Al saturation of the crops grown. Lime requirement (LR) was quantified using soil dataset from Africa soil information service (AfSIS). Spatial variations of LR of cereals, pulses and cash crops were predicted using random forest algorithm. Our results revealed that mean of LR Mg CaCO3 (1 Mg = 106 g) ha−1 for cereal crops were 6.34, 6.35, and 4.41 for maize, sorghum, and upland rice, respectively. Mean of LR (Mg ha−1) for pulses were 6.28, 5.19, and 4.90 for common beans, soybeans, and cowpeas, respectively. Mean of LR Mg CaCO3 (1 Mg = 106 g) ha−1 for cash crops were 3.41 and 6.29 for coffee and cotton, respectively. Spatial variation showed that LR in croplands was higher in tropical humid regions than in semi-arid and arid regions and ranged from 0 to 8.8 Mg ha−1. The results of 10-fold cross validation for high model performance of LR for tested crops were coefficient of determination (R2) of 0.61, a root mean square error (RMSE) of 0.5, and a mean absolute error (MAE) of 0.31, maize LR with RMSE = 0.9, MAE = 0.24, and R2 = 0.51, and cotton LR with RMSE = 0.5, MAE = 0.31, and R2 = 0.60. We recommend predicting lime requirement in acidic soils of Sub-Saharan Africa by adjusting Al saturation up to the tolerance of the grown crop, updating soil surveys in Sab Saharan Africa, and using digital soil mapping to monitor soil acidity and lime requirement.

MP29-09 IDENTIFICATION OF PROSTATE CANCER METASTATIC DISEASE FOR DIFFERENT RISK GROUPS BASED ON FASTTEXT WORD EMBEDDING AND SUPERVISED LEARNING

You have accessJournal of UrologyCME1 Apr 2023MP29-09 IDENTIFICATION OF PROSTATE CANCER METASTATIC DISEASE FOR DIFFERENT RISK GROUPS BASED ON FASTTEXT WORD EMBEDDING AND SUPERVISED LEARNING Ruixin Yang, Michael Burns, Amanda De Hoedt, Stephen Williams, Stephen Freedland, and Zachary Klaassen Ruixin YangRuixin Yang More articles by this author , Michael BurnsMichael Burns More articles by this author , Amanda De HoedtAmanda De Hoedt More articles by this author , Stephen WilliamsStephen Williams More articles by this author , Stephen FreedlandStephen Freedland More articles by this author , and Zachary KlaassenZachary Klaassen More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003257.09AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Prostate Cancer (PCa) is typically associated with metastases to bone and/or soft-tissue. Many studies have observed a survival benefit in timely identification of metastatic prostate cancer (mPCa). We and others have recently published natural language processing (NLP) tools to identify metastatic disease. We aimed to build upon our prior work to develop a FastText based supervised learning model for rapid mPCa identification through text data of individual scans. METHODS: We collected 128,092 scans on 6,959 unique patients from the U.S. Veterans Affairs Health System. We built a supervised learning model based on FastText word embedding and evaluated its performance for different patient risk groups , and where ICD codes for metastasis, CRPC, receipt of certain drugs and ADT were taken into consideration. We finally reported the statistical results of 10-fold cross validation against hand-abstracted (i.e. “true”) metastasis status. RESULTS: Overall, among 6,959 patients, the model had an avg AUC of 0.96 (averaged over 10-fold cross validation), with high sensitivity and specificity for predicting a positive result (85% and 93%, respectively) when 0.5 cut-point was selected. PPV was 85%, NPV was 92% and F1 score was 0.85. For different risk-groups (N=4,739), the model had high AUCs ranged from 0.91 to 0.94 depending on the clinical characteristics of the cohorts used in the validation. The avg model training time for the training set (115k notes on 6.3k patients) was < 5 mins and the avg identification time for validation set (13k notes on 700 patients) notes was only 1s. CONCLUSIONS: This study systematically investigated the programmatic identification of mPCa in a large cohort using word embedding and supervised learning and evaluated the model performance in different risk groups. Overall, the model had extremely high accuracy and higher than our previously published model. In addition, our method is extremely fast either for model training or final identification. Thus, this study provides a practical tool for rapid and accurately identifying patients with metastases in the VA system, an essential step forward in population-based research. Source of Funding: Merck © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e384 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information Ruixin Yang More articles by this author Michael Burns More articles by this author Amanda De Hoedt More articles by this author Stephen Williams More articles by this author Stephen Freedland More articles by this author Zachary Klaassen More articles by this author Expand All Advertisement PDF downloadLoading ...

ELMo4m6A: A Contextual Language Embedding-Based Predictor for Detecting RNA N6-Methyladenosine Sites.

N6-methyladenosine (m6A) is a universal post-transcriptional modification of RNAs, and it is widely involved in various biological processes. Identifying m6A modification sites accurately is indispensable to further investigate m6A-mediated biological functions. How to better represent RNA sequences is crucial for building effective computational methods for detecting m6A modification sites. However, traditional encoding methods require complex biological prior knowledge and are time-consuming. Furthermore, most of the existing m6A sites prediction methods are limited to single species, and few methods are able to predict m6A sites across different species and tissues. Thus, it is necessary to design a more efficient computational method to predict m6A sites across multiple species and tissues. In this paper, we proposed ELMo4m6A, a contextual language embedding-based method for predicting m6A sites from RNA sequences without any prior knowledge. ELMo4m6A first learns embeddings of RNA sequences using a language model ELMo, then uses a hybrid convolutional neural network (CNN) and long short-term memory (LSTM) to identify m6A sites. The results of 5-fold cross-validation and independent testing demonstrate that ELMo4m6A is superior to state-of-the-art methods. Moreover, we applied integrated gradients to find potential sequence patterns contributing to m6A sites.

Early estimation of the long-term deflection of reinforced concrete beams using surrogate models

This paper describes the development and testing of a novel artificial intelligence-based inference model for the early prediction of long-term deflection in RC beams, which is a critical but onerous task for civil engineers. This model, called WFR-FBI-LSSVR, integrates wrapper-based feature refinement (WFR) with the forensic-based investigation (FBI) algorithm and the least squares support vector regression (LSSVR) technique. In this model, the FBI algorithm performs the optimization process to gradually fine-tune LSSVR’s hyper-parameter values and wrapper-refined features, while LSSVR uses the received potential hyper-parameter values (γ and C) and sets of features to build comparative inference models, which are used to generate prediction values that are then used to calculate the objective function values of FBI. A comprehensive survey of the literature collected data on a wide range of settings to improve the construction of WFR-FBI-LSSVR and verify its performance. The statistical results of 10-fold cross-validation confirmed WFR-FBI-LSSVR as significantly more accurate than several widely used AI models and mathematical approaches, achieving the best values in terms of root mean square error (RMSE = 7.86 mm), mean absolute percentage error (MAPE = 15.21 %), and coefficient of determination (R2 = 0.908). The calculated prediction-to-actual value ratio further validated the robustness of WFR-FBI-LSSVR with an average ratio value of 1.01, which was the closest value to 1 achieved among the models and approaches compared in this study. In summary, this paper contributes a novel free-formulae prediction model for researchers and civil engineers to accurately predict long-term deflection in RC beams.

Cross-subject classification of depression by using multiparadigm EEG feature fusion

Background and objectiveThe aim of this study is to solve the non-stationarity and complexity characteristics and huge individual differences in the electroencephalogram (EEG) signals for depression classification. MethodsTo address those problems, the Lempel–Ziv complexity feature matrices were extracted from the EEG signals under the two paradigms of eyes open and eyes closed in the resting state. Topographical map of brain and statistical analysis were introduced to investigate the significance of eyes open and eyes closed EEG for depression classification. To promote the classification accuracy, feature matrices from the two paradigms were fused. And linear combination and concatenation fusion methods were proposed to further reveal the underlying mechanism of improving classification accuracy. Support vector machine (SVM), K-nearest neighbor, and decision tree classifiers were employed and compared to classify depression under the eyes open, eyes closed and fused paradigm. ResultsThe classification results of 10-fold cross-validation showed that the highest average accuracy (86.58%) under a single paradigm was achieved in the eyes-open state. The multiparadigm fusion method of concatenation was better than the linear combination. The best classification result was obtained using multiparadigm feature concatenation under the SVM classifier, yielding an accuracy of 94.03%. ConclusionThe multiparadigm feature fusion method proposed in this paper can effectively improve the accuracy of depression classification. It was proved that eyes open and eyes closed EEG have complementary information, which was benefit for the cross-subject classification of depression. It provides new ideas for depression classification in clinics.

Graph regularized non-negative matrix factorization with prior knowledge consistency constraint for drug–target interactions prediction

BackgroundIdentifying drug–target interactions (DTIs) plays a key role in drug development. Traditional wet experiments to identify DTIs are expensive and time consuming. Effective computational methods to predict DTIs are useful to narrow the searching scope of potential drugs and speed up the process of drug discovery. There are a variety of non-negativity matrix factorization based methods to predict DTIs, but the convergence of the algorithms used in the matrix factorization are often overlooked and the results can be further improved.ResultsIn order to predict DTIs more accurately and quickly, we propose an alternating direction algorithm to solve graph regularized non-negative matrix factorization with prior knowledge consistency constraint (ADA-GRMFC). Based on known DTIs, drug chemical structures and target sequences, ADA-GRMFC at first constructs a DTI matrix, a drug similarity matrix and a target similarity matrix. Then DTI prediction is modeled as the non-negative factorization of the DTI matrix with graph dual regularization terms and a prior knowledge consistency constraint. The graph dual regularization terms are used to integrate the information from the drug similarity matrix and the target similarity matrix, and the prior knowledge consistency constraint is used to ensure the matrix decomposition result should be consistent with the prior knowledge of known DTIs. Finally, an alternating direction algorithm is used to solve the matrix factorization. Furthermore, we prove that the algorithm can converge to a stationary point. Extensive experimental results of 10-fold cross-validation show that ADA-GRMFC has better performance than other state-of-the-art methods. In the case study, ADA-GRMFC is also used to predict the targets interacting with the drug olanzapine, and all of the 10 highest-scoring targets have been accurately predicted. In predicting drug interactions with target estrogen receptors alpha, 17 of the 20 highest-scoring drugs have been validated.

A 70‑RNA model based on SVR and RFE for predicting the pancreatic cancer clinical prognosis.

Researches on the prognosis of pancreatic cancer is of great significance to improve the patient treatment effect and survival. Current researches mainly focus on the prediction of the survival status and the determination of prognostic markers. Each patient has its own characteristics, there is no report about the prediction of survival time. However, accurate prediction of survival time is critical for personalized medicine. In this paper, a hybrid algorithm of Support Vector Regression (SVR) and Recursive Feature Elimination (RFE) was used to construct a quantitative prediction model of Overall Survival (OS) for pancreatic cancer patients, 70 RNAs related to OS were determined, including 33 mRNAs, 28 lncRNAs, and 9 miRNAs. The results of 10-fold cross-validation (R2 is 0.9693) and the generalization ability (R2 is 0.9666) showed that the model has reliable predictive performance and these 70 RNAs are important factors influencing the OS of pancreatic cancer patients. To further study the relationship between RNA-RNA interaction and the survival, competitive endogenous RNA (ceRNA) regulation network was constructed. Degree centrality, betweenness centrality and closeness centrality of nodes in the ceRNA network showed that hsa-mir-570, hsa-mir-944, hsa-mir-6506, hsa-mir-3136, MMP16, PLGLB2, HPGD, FUT1, MFSD2A, SULT1E1, SLC13A5, ZNF488, F2RL2, TNFRSF8, TNFSF11, FHDC1, ISLR2 and THSD7B are hub nodes, which are key RNAs closely determining the OS of pancreatic cancer patients.

Prediction of adolescent suicidal ideation after the COVID-19 pandemic: A nationwide survey of a representative sample of Korea.

ObjectiveThis study developed a model to predict groups vulnerable to suicidal ideation after the declaration of the COVID-19 pandemic based on nomogram techniques targeting 54,948 adolescents who participated in a national survey in South Korea.MethodsThis study developed a model to predict suicidal ideation by using logistic regression analysis. The model aimed to understand the relationship between predictors associated with the suicidal ideation of South Korean adolescents by using the top seven variables with the highest feature importance confirmed in XGBoost (extreme gradient boosting). The regression model was developed using a nomogram so that medical workers could easily interpret the probability of suicidal ideation and identify groups vulnerable to suicidal ideation.ResultsThis epidemiological study predicted that eighth graders who experienced depression in the past 12 months, had a lot of subjective stress, frequently felt lonely in the last 12 months, experienced much-worsened household economic status during the COVID-19 pandemic, and had poor academic performance were vulnerable to suicidal ideation. The results of 10-fold cross-validation revealed that the area under the curve (AUC) of the adolescent suicidal ideation prediction nomogram was 0.86, general accuracy was 0.89, precision was 0.87, recall was 0.89, and the F1-score was 0.88.ConclusionIt is required to recognize the seriousness of adolescent suicide and mental health after the onset of the COVID-19 pandemic and prepare a customized support system that considers the characteristics of persons at risk of suicide at the school or community level.

Application of Multivariate Adaptive Regression Splines (MARS) approach in prediction of compressive strength of eco-friendly concrete

Concrete is the most often used material in the building sector. The use of ground-granulated blast-furnace slag (GGBFS) and recycled concrete aggregates (RCA) in concrete reduces its negative environmental impact. However, increasing the number of ingredients in concrete complicates the prediction of its compressive strength. This research aims to develop a machine learning model for predicting the compressive strength of eco-friendly concrete using Multivariate Adaptive Regression Splines (MARS). MARS is a well-known technique for creating predictive modeling equations using experimental data. A data set of 161 concrete specimens was obtained to train and evaluate the machine learning algorithms for this work. The findings of the MARS prediction algorithm were compared to those of the Support Vector Machine and the Random Forest regression algorithms (black-box models) using the training set. A hyperparameter tuning procedure was applied to find the optimal MARS, RF, and SVM model parameters. The results of 5-fold cross-validation indicate that the MARS optimal model has the highest coefficient of determination (R2), with a value equal to 0.889, and the lowest Root Mean Squared Error (RMSE) measure, with a value equal to 4.110 MPa. For our case, the performance metrics measures proved the superiority of the MARS model when compared with RF and SVM. The final MARS optimal prediction equations were built using a full training set and were then evaluated using unseen data (testing set). Our model has the potential to help civil engineers in designing durable infrastructures.

PSuc-FFSEA: Predicting Lysine Succinylation Sites in Proteins Based on Feature Fusion and Stacking Ensemble Algorithm.

Being a new type of widespread protein post-translational modifications discovered in recent years, succinylation plays a key role in protein conformational regulation and cellular function regulation. Numerous studies have shown that succinylation modifications are closely associated with the development of many diseases. In order to gain insight into the mechanism of succinylation, it is vital to identify lysine succinylation sites. However, experimental identification of succinylation sites is time-consuming and laborious, and traditional identification tools are unable to meet the rapid growth of datasets. Therefore, to solve this problem, we developed a new predictor named pSuc-FFSEA, which can predict succinylation sites in protein sequences by feature fusion and stacking ensemble algorithm. Specifically, the sequence information and physicochemical properties were first extracted using EBGW, One-Hot, continuous bag-of-words, chaos game representation, and AAF_DWT. Following that, feature selection was performed, which applied LASSO to select the optimal subset of features for the classifier, and then, stacking ensemble classifier was designed using two-layer stacking ensemble, selecting three classifiers, SVM, broad learning system and LightGBM classifier, as the base classifiers of the first layer, using logistic regression classifier as the meta classifier of the second layer. In order to further improve the model prediction accuracy and reduce the computational effort, bayesian optimization algorithm and grid search algorithm were utilized to optimize the hyperparameters of the classifier. Finally, the results of rigorous 10-fold cross-validation indicated our predictor showed excellent robustness and performed better than the previous prediction tools, which achieved an average prediction accuracy of 0.7773 ± 0.0120. Besides, for the convenience of the most experimental scientists, a user-friendly and comprehensive web-server for pSuc-FFSEA has been established at https://bio.cangmang.xyz/pSuc-FFSEA, by which one can easily obtain the expected data and results without going through the complicated mathematics.

Comprehensive Prediction of Lipocalin Proteins Using Artificial Intelligence Strategy.

Lipocalin belongs to the calcyin family, and its sequence length is generally between 165 and 200 residues. They are mainly stable and multifunctional extracellular proteins. Lipocalin plays an important role in several stress responses and allergic inflammations. Because the accurate identification of lipocalins could provide significant evidences for the study of their function, it is necessary to develop a machine learning-based model to recognize lipocalin. In this study, we constructed a prediction model to identify lipocalin. Their sequences were encoded by six types of features, namely amino acid composition (AAC), composition of k-spaced amino acid pairs (CKSAAP), pseudo amino acid composition (PseAAC), Geary correlation (GD), normalized Moreau-Broto autocorrelation (NMBroto) and composition/transition/distribution (CTD). Subsequently, these features were optimized by using feature selection techniques. A classifier based on random forest was trained according to the optimal features. The results of 10-fold cross-validation showed that our computational model would classify lipocalins with accuracy of 95.03% and area under the curve of 0.987. On the independent dataset, our computational model could produce the accuracy of 89.90% which was 4.17% higher than the existing model. In this work, we developed an advanced computational model to discriminate lipocalin proteins from non-lipocalin proteins. In the proposed model, protein sequences were encoded by six descriptors. Then, feature selection was performed to pick out the best features which could produce the maximum accuracy. On the basis of the best feature subset, the RF-based classifier can obtained the best prediction results.

Identification of DNA modification sites based on elastic net and bidirectional gated recurrent unit with convolutional neural network

DNA N4-methylcytosine (4mC) and DNA N6-methyladenine (6mA) are significant epigenetic modifications. 4mC is closely related to the restriction modification system, and 6mA has a hand in the process of various cellular activities. In order to further explore their functional mechanisms and biological significance, and to overcome the bottleneck of narrow coverage in traditional experimental methods, it is needed to propose an efficient prediction method with a wide range of applications. In this work, we develop a prediction method named 4mCi6mA-BGC to predict 4mC sites and 6mA sites. First, we employ binary, K-mer nucleotide frequency (K-mer), pseudo K-tuple nucleotide composition (PseKNC), dinucleotide-based auto covariance (DAC) and monoDiKGap theoretical description (MonoDiKGap) to encode DNA sequences. Then, the elastic net is employed for feature selection, and the optimized feature space is put into a deep learning framework composed of bidirectional gated recurrent unit and convolutional neural network. The benchmark datasets include six datasets, which contain 14 328 4mC sites from different species. The results of 10-fold cross-validation indicate that the prediction accuracy significantly outperforms the existing prediction methods. Meanwhile, use independent datasets Rice and Arabidopsis thaliana to further confirm the predictive ability of 4mCi6mA-BGC. Compared with the existing prediction methods, 4mCi6mA-BGC shows the best prediction performance. These comprehensive results indicate that our method can identify DNA modification sites represented by 4mC and 6mA sites.

Developing a nomogram for predicting the depression of senior citizens living alone while focusing on perceived social support.

BACKGROUNDAlthough the number of senior citizens living alone is increasing, only a few studies have identified factors related to the depression characteristics of senior citizens living alone by using epidemiological survey data that can represent a population group.AIMTo evaluate prediction performance by building models for predicting the depression of senior citizens living alone that included subjective social isolation and perceived social support as well as personal characteristics such as age and drinking.METHODSThis study analyzed 1558 senior citizens (695 males and 863 females) who were 60 years or older and completed an epidemiological survey representing the South Korean population. Depression, an outcome variable, was measured using the short form of the Korean version CES-D (short form of CES-D).RESULTSThe prevalence of depression among the senior citizens living alone was 7.7%. The results of multiple logistic regression analysis showed that the experience of suicidal urge over the past year, subjective satisfaction with help from neighbors, subjective loneliness, age, and self-esteem were significantly related to the depression of senior citizens living alone (P < 0.05). The results of 10-fold cross validation showed that the area under the curve of the nomogram was 0.96, and the F1 score of it was 0.97. CONCLUSIONIt is necessary to strengthen the social network of senior citizens living alone with friends and neighbors based on the results of this study to protect them from depression.