Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective To explore the feasibility of long non-coding RNAs (lncRNAs) as biomarkers for radiation-induced intestinal injury. Methods Mice were exposed to 15 Gy of ⁶⁰Co γ-rays to the abdominal area. The pathological changes in intestinal tissues were analyzed at 72 h post-irradiation to confirm the successful establishment of the radiation-induced intestinal injury model. Real-time quantitative PCR was conducted to detect the expression of candidate radiation-responsive lncRNAs in the jejunum, jejunal crypts, colon tissues, and plasma of irradiated mice. Human intestinal epithelial cell line HIEC-6 and human colon epithelial cell line NCM460 were exposed to 0, 5, 10, and 15 Gy of ⁶⁰Co γ-rays. The expression levels of candidate lncRNAs were measured at 4, 24, 48, and 72 h post-irradiation to observe their changes with the irradiation dose. Results Pathological analysis showed that abdominal irradiation with 15 Gy successfully established an acute radiation-induced intestinal injury mouse model. Real-time quantitative PCR showed that Dino, Lncpint, Meg3, Dnm3os, Trp53cor1, Pvt1, and Neat1 were significantly upregulated following the occurrence of radiation-induced intestinal injury (P < 0.05). Among them, Meg3 and Dnm3os in mouse plasma were significantly upregulated (P < 0.05), while Gas5 was significantly downregulated (P < 0.05). In HIEC-6 and NCM460 cells, the expression levels of DINO, MEG3, DNM3OS, and GAS5 showed dose-dependent patterns at certain time points (P < 0.05). Conclusion The lncRNAs encoded by MEG3, DNM3OS, and GAS5 in intestinal epithelial cells are responsive to ionizing radiation. Consistent differential expression changes were detected in mouse plasma and intestinal tissues, indicating their potential as biomarkers for radiation-induced intestinal injury.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Humans ; Tranexamic Acid/therapeutic use* ; Arthroplasty, Replacement, Knee ; Antifibrinolytic Agents/therapeutic use* ; Blood Loss, Surgical ; Postoperative Hemorrhage

Natural products are important sources for the discovery of anti-tumor drugs. Evodiamine is the main alkaloid component of the traditional Chinese herb Wu-Chu-Yu, and it has weak antitumor activity. In recent years, a number of highly active antitumor candidates have been discovered with a significant progress. This article reviews the research progress of evodiamine-based antitumor drug design strategies, in order to provide reference for the development of new drugs with natural products as leads.

The analysis of ammonia nitrogen in real water samples is challenging due to matrix interferences and difficulties for rapid on-site analysis.On the basis of the standard method,i.e.water quality-determination of ammonia nitrogen-salicylic acid spectrophotometry(HJ 536-2009),a simple device for online detecting ammonia nitrogen was developed using a sequential injection analysis(SIA)system in this work.The ammonia nitrogen transformation system,color reaction system,and detection system were built in compatible with the SIA system,respectively.In particular,the detection system was assembled by employing light-emitting diode as the light source,photodiode as the detector,and polyvinylchloride tube as the cuvette,thus significantly reducing the volume,energy consumption and fabricating cost of the detection system.As a result,the accurate analysis of ammonia nitrogen in complex water samples was achieved.A quantitative detection of ammonia nitrogen in water sample was obtained in 12 min,along with linear range extending to 1000 μmol/L,precisions(Relative standard deviation,RSD)of 4.3%(C=10 μmol/L,n=7)and 4.2%(C=500 μmol/L,n=7),and limit of detection(LOD)of 0.65 μmol/L(S/N=3,n=7).The results of interfering experiments showed that the detection of ammonia nitrogen by the developed device was not interfered by the common coexisting ions and components,therefore the environmental water could be directly analyzed,such as reservoir water,domestic sewage,sea water and leachate of waste landfill.The analytical results were consistent with those obtained by the environmental protection standard method(Water quality determination of ammonia nitrogen-salicylic acid spectrophotometry,HJ 536-2009).In addition,the spiking recoveries were in the range of 92.3%-98.1%,further confirming the accuracy and practicality of the developed device.

Objective:To study the current status of longitudinal extrauterine growth restriction (EUGR) in extremely preterm infants (EPIs) and to develop a prediction model based on clinical data from multiple NICUs.Methods:From January 2017 to December 2018, EPIs admitted to 32 NICUs in North China were retrospectively studied. Their general conditions, nutritional support, complications during hospitalization and weight changes were reviewed. Weight loss between birth and discharge > 1SD was defined as longitudinal EUGR. The EPIs were assigned into longitudinal EUGR group and non-EUGR group and their nutritional support and weight changes were compared. The EPIs were randomly assigned into the training dataset and the validation dataset with a ratio of 7∶3. Univariate Cox regression analysis and multiple regression analysis were used in the training dataset to select the independent predictive factors. The best-fitting Nomogram model predicting longitudinal EUGR was established based on Akaike Information Criterion. The model was evaluated for discrimination efficacy, calibration and clinical decision curve analysis.Results:A total of 436 EPIs were included in this study, with a mean gestational age of (26.9±0.9) weeks and a birth weight of (989±171) g. The incidence of longitudinal EUGR was 82.3%(359/436). Seven variables (birth weight Z-score, weight loss, weight growth velocity, the proportion of breast milk ≥75% within 3 d before discharge, invasive mechanical ventilation ≥7 d, maternal antenatal corticosteroids use and bronchopulmonary dysplasia) were selected to establish the prediction model. The area under the receiver operating characteristic curve of the training dataset and the validation dataset were 0.870 (95% CI 0.820-0.920) and 0.879 (95% CI 0.815-0.942), suggesting good discrimination efficacy. The calibration curve indicated a good fit of the model ( P>0.05). The decision curve analysis showed positive net benefits at all thresholds. Conclusions:Currently, EPIs have a high incidence of longitudinal EUGR. The prediction model is helpful for early identification and intervention for EPIs with higher risks of longitudinal EUGR. It is necessary to expand the sample size and conduct prospective studies to optimize and validate the prediction model in the future.