Cross-validation Data Research Articles

Investigating ocean surface responses to typhoons using reconstructed satellite data

Due to the cloud blocking, the research on the responses of upper ocean to typhoons using satellite data is greatly affected. At present, most of the solutions are based on lower time resolution data or composite satellite data. In this paper, multi-source satellite data were fused by weighted average method and reconstructed through the Data INterpolating Convolutional Auto-Encoder method (DINCAE) and used to investigate daily sea surface responses to typhoons. The root mean squared error (RMSE) between the reconstructed data and the cross-validation data for SST and CHL was only 1.093 °C and 0.7833 mg/m3. Three response patterns were reported over the East China Sea (ECS). The first one was that both SST and CHL have significant changes, the second was that only SST changes a lot, and the third one was only CHL changes significantly after typhoons. Moreover, it was found that, after some typhoons, the SST or CHL had a wide range of changes, but the amount of changes is not large. The different responses were related to the wind speed and moving speed of typhoons. In addition, when two typhoons pass by successively, although the wind speed is low, the sea surface would have significant responses. Further, areas of significant SST drops coincide with negative sea level anomaly (SLA) fields in some typhoon tracks. SST drops to its minimum level 2 days after the arrival of a typhoon and returns to its previous level 15 days after the arrival. CHL keeps increasing until 4 days after the typhoon’s passage, and then drops to its initial level around 10 days after the passage. The responses of SST and CHL decrease with distance, with prominent changes occurring within 100 km of the typhoon’s center.

KLASIFIKASI KEJADIAN HIPERTENSI DENGAN METODE SUPPORT VECTOR MACHINE (SVM) MENGGUNAKAN DATA PUSKESMAS DI KOTA BANDAR LAMPUNG

Hypertension is a condition in which a person experiences an increase in blood pressure above the normal value which causes pain and even death. Normal human blood pressure is 120/80 mmHg. Patients with hypertension cannot be cured, but prevention and control can be done. The hypertension cases are always increasing in Indonesia. The Bandar Lampung City Health Service stated that hypertension is a disease that always ranks in the top ten diseases in Bandar Lampung City. Diagnosis of hypertension is currently manually performed by requiring a lot of energy, materials, and time. Based on the condition, there is an idea to apply the field of biomedical data analysis to help diagnosing hypertension using the support vector machine (SVM) method in Bandar Lampung City. This study classifies and measures the accuracy of the support vector machine method in hypertension. The data comes from five health centers in Bandar Lampung City from 2017 to 2019 with 10-fold cross validation data sharing. The kernels used are linear, gaussian, and polynomial kernels. This study successfully classifies hypertension sufferers in Bandar Lampung City. The result of the highest feature correlation analysis is 0.90. The results of the classification using the support vector machine method get the highest accuracy, which is 99.78% on the gaussian kernel.

A novel framework for designing a multi-DoF prosthetic wrist control using machine learning

Prosthetic arms can significantly increase the upper limb function of individuals with upper limb loss, however despite the development of various multi-DoF prosthetic arms the rate of prosthesis abandonment is still high. One of the major challenges is to design a multi-DoF controller that has high precision, robustness, and intuitiveness for daily use. The present study demonstrates a novel framework for developing a controller leveraging machine learning algorithms and movement synergies to implement natural control of a 2-DoF prosthetic wrist for activities of daily living (ADL). The data was collected during ADL tasks of ten individuals with a wrist brace emulating the absence of wrist function. Using this data, the neural network classifies the movement and then random forest regression computes the desired velocity of the prosthetic wrist. The models were trained/tested with ADLs where their robustness was tested using cross-validation and holdout data sets. The proposed framework demonstrated high accuracy (F-1 score of 99% for the classifier and Pearson’s correlation of 0.98 for the regression). Additionally, the interpretable nature of random forest regression was used to verify the targeted movement synergies. The present work provides a novel and effective framework to develop an intuitive control for multi-DoF prosthetic devices.

Predicting Depression From Smartphone Behavioral Markers Using Machine Learning Methods, Hyperparameter Optimization, and Feature Importance Analysis: Exploratory Study.

BackgroundDepression is a prevalent mental health challenge. Current depression assessment methods using self-reported and clinician-administered questionnaires have limitations. Instrumenting smartphones to passively and continuously collect moment-by-moment data sets to quantify human behaviors has the potential to augment current depression assessment methods for early diagnosis, scalable, and longitudinal monitoring of depression.ObjectiveThe objective of this study was to investigate the feasibility of predicting depression with human behaviors quantified from smartphone data sets, and to identify behaviors that can influence depression.MethodsSmartphone data sets and self-reported 8-item Patient Health Questionnaire (PHQ-8) depression assessments were collected from 629 participants in an exploratory longitudinal study over an average of 22.1 days (SD 17.90; range 8-86). We quantified 22 regularity, entropy, and SD behavioral markers from the smartphone data. We explored the relationship between the behavioral features and depression using correlation and bivariate linear mixed models (LMMs). We leveraged 5 supervised machine learning (ML) algorithms with hyperparameter optimization, nested cross-validation, and imbalanced data handling to predict depression. Finally, with the permutation importance method, we identified influential behavioral markers in predicting depression.ResultsOf the 629 participants from at least 56 countries, 69 (10.97%) were females, 546 (86.8%) were males, and 14 (2.2%) were nonbinary. Participants’ age distribution is as follows: 73/629 (11.6%) were aged between 18 and 24, 204/629 (32.4%) were aged between 25 and 34, 156/629 (24.8%) were aged between 35 and 44, 166/629 (26.4%) were aged between 45 and 64, and 30/629 (4.8%) were aged 65 years and over. Of the 1374 PHQ-8 assessments, 1143 (83.19%) responses were nondepressed scores (PHQ-8 score <10), while 231 (16.81%) were depressed scores (PHQ-8 score ≥10), as identified based on PHQ-8 cut-off. A significant positive Pearson correlation was found between screen status–normalized entropy and depression (r=0.14, P<.001). LMM demonstrates an intraclass correlation of 0.7584 and a significant positive association between screen status–normalized entropy and depression (β=.48, P=.03). The best ML algorithms achieved the following metrics: precision, 85.55%-92.51%; recall, 92.19%-95.56%; F1, 88.73%-94.00%; area under the curve receiver operating characteristic, 94.69%-99.06%; Cohen κ, 86.61%-92.90%; and accuracy, 96.44%-98.14%. Including age group and gender as predictors improved the ML performances. Screen and internet connectivity features were the most influential in predicting depression.ConclusionsOur findings demonstrate that behavioral markers indicative of depression can be unobtrusively identified from smartphone sensors’ data. Traditional assessment of depression can be augmented with behavioral markers from smartphones for depression diagnosis and monitoring.

Glocalizing ELT Reform in China: A Perspective from the Use of English in the Workplace

With the largest number of English learners in the world, the influence of the English language teaching (ELT) reform in China cannot be underestimated. This article explores the implications of the actual use of English in China’s workplace for ELT reform in the context of English as a lingua franca (ELF). On the basis of cross-validated data (questionnaire survey and focused interview) collected from 2495 participants, we argue that ELT reform in China should be geared towards using English communicatively in ELF settings –that is, firstly, ELT curriculum and pedagogies should focus more broadly on improving students’ communication skills instead of narrowly measuring whether they have successfully adhered to lexico-grammatical accuracy pertaining to Standard English norms. Secondly, the native-speaker-based pedagogical model of ELT in China should be enriched with judiciously selected indigenized variants as long as meaning is not adversely affected. Last but not least, for ELT reforms to bear fruit, it is absolutely crucial to ensure a steady supply of properly trained and resourceful ELT teachers.

Prediction of the motion of chest internal points using a recurrent neural network trained with real-time recurrent learning for latency compensation in lung cancer radiotherapy.

During the radiotherapy treatment of patients with lung cancer, the radiation delivered to healthy tissue around the tumor needs to be minimized, which is difficult because of respiratory motion and the latency of linear accelerator (LINAC) systems. In the proposed study, we first use the Lucas-Kanade pyramidal optical flow algorithm to perform deformable image registration (DIR) of chest computed tomography (CT) scan images of four patients with lung cancer. We then track three internal points close to the lung tumor based on the previously computed deformation field and predict their position with a recurrent neural network (RNN) trained using real-time recurrent learning (RTRL) and gradient clipping. The breathing data is quite regular, sampled at approximately 2.5 Hz, and includes artificially added drift in the spine direction. The amplitude of the motion of the tracked points ranged from 12.0 mm to 22.7 mm. Finally, we propose a simple method for recovering and predicting three-dimensional (3D) tumor images from the tracked points and the initial tumor image, based on a linear correspondence model and the Nadaraya-Watson non-linear regression. The root-mean-square (RMS) error, maximum error and jitter corresponding to the RNN prediction on the test set were smaller than the same performance measures obtained with linear prediction and least mean squares (LMS). In particular, the maximum prediction error associated with the RNN, equal to 1.51 mm, is respectively 16.1% and 5.0% lower than the error given by a linear predictor and LMS. The average prediction time per time step with RTRL is equal to 119 ms, which is less than the 400 ms marker position sampling time. The tumor position in the predicted images appears visually correct, which is confirmed by the high mean cross-correlation between the original and predicted images, equal to 0.955. The standard deviation of the Gaussian kernel and the number of layers in the optical flow algorithm were the parameters having the most significant impact on registration performance. Their optimization led respectively to a 31.3% and 36.2% decrease in the registration error. Using only a single layer proved to be detrimental to the registration quality because tissue motion in the lower part of the lung has a high amplitude relative to the resolution of the CT scan images. The random initialization of the hidden units and the number of these hidden units were found to be the most important factors affecting the performance of the RNN. Increasing the number of hidden units from 15 to 250 led to a 56.3% decrease in the prediction error on the cross-validation data. Similarly, optimizing the standard deviation of the initial Gaussian distribution of the synaptic weights σinitRNN led to a 28.4% decrease in the prediction error on the cross-validation data, with the error minimized for σinitRNN=0.02 with the four patients.

Bayesian ridge regression shows the best fit for SSR markers in Psidium guajava among Bayesian models

Markers are an important tool in plant breeding, which can improve conventional phenotypic breeding, generating more accurate information outcoming better decision making. This study aimed to apply and compare the fit of different Bayesian models BRR, BayesA, BayesB, BayesB (setting the value from very low to pi = {10}^{-5}), BayesC and Bayesian Lasso (LASSO) for predictions of the genomic genetic values of productivity and quality traits of a guava population. The models were fitted for traits fruit mass, pulp mass, soluble solids content, fruit number, and production per plant in the genomic prediction with SSR markers, obtained through the CTAB extraction method with 200 primers. The Bayesian ridge regression model showed the best results for all traits and was chosen to predict the individual’s genomic values according to the cross-validation data. A good stabilization of the Markov and Monte Carlo chains was observed with the mean values close to the observed phenotypic means. Heritabilities showed good predictive accuracy. The model showed strong correlations between some traits, allowing indirect selection.

Predicting male dairy calf live weight for use in calf management decision support

The growing awareness and scrutiny of the management of young dairy calves, especially male calves, necessitates a support tool to aid in the planning of resource allocation on dairy farms. There is a desire among some vendors for a minimum calf weight when purchasing young dairy bull calves. Hence, the objective of the present study was to investigate whether live weight of young calves (approximately 10–50 d old) can be predicted using readily accessible animal-level features, especially features that may be available in advance of birth. A multiple linear regression mixed model was developed with the live weight of 602 dairy bull calves aged between 10 and 42 d as the dependent variable; the age at which an animal is predicted to reach a predefined live weight was then estimated based on the model regression coefficients. Fixed effects included in the multiple regression model were dam parity, gestation length, and parental average genetic merit for relevant traits available in Ireland; namely, birth weight, birth size, and carcass weight. Herd of origin was included as a random effect, with all calves having been sold directly from the farm of birth. Live weight data were recorded at the point of sale when calves were, on average, 26 d old with a mean live weight of 56.6 kg. Animals were randomly assigned to 10 separate (i.e., folds) cross-validation data sets without replacement (i.e., each fold consisted of a different 10% of the data to test the model, with the remaining 90% of data being used to train the model) to quantify the accuracy of prediction. Across all data, the correlation between actual and predicted live weight was 0.76; the regression coefficient of actual live weight on predicted live weight across all data was 0.99. The root mean squared error of prediction varied from 4.40 to 6.66 kg per fold. Across all data, the root mean squared error was 5.61 kg, implying that 68% of live weight predictions were within 5.61 kg of the actual live weight. Given the potential availability of all model features in advance of birth (gestation length can be predicted from ultrasound examination of the pregnant uterus, although substituting parental average genetic merit for gestation length had minimal effect on model performance), predictions can be integrated into a dairy farm decision support tool to aid in the management of labor and infrastructure resources to achieve minimum live weight specifications before sale.

Comparison of genetic risk prediction models to improve prediction of coronary heart disease in two large cohorts of the MONICA/KORA study.

It is still unclear how genetic information, provided as single-nucleotide polymorphisms (SNPs), can be most effectively integrated into risk prediction models for coronary heart disease (CHD)to add significant predictive value beyond clinical risk models. For the present study, a population-based case-cohort was used as a trainingset (451 incident cases, 1488 noncases) and an independent cohort as testset (160 incident cases, 2749 noncases). The following strategies to quantify genetic information were compared: A weighted genetic risk score including Metabochip SNPs associated with CHD in the literature (GRSMetabo ); selection of the most predictive SNPs among these literature-confirmed variants using priority-Lasso (PLMetabo ); validation of two comprehensive polygenic risk scores: GRSGola based on Metabochip data, and GRSKhera (available in the testset only) based on cross-validated genome-wide genotyping data. We used Cox regression to assess associations with incident CHD. C-index, category-free net reclassification index (cfNRI) and relative integrated discrimination improvement (IDIrel ) were used to quantify the predictive performance of genetic information beyond Framingham risk score variables. In contrast to GRSMetabo and PLMetabo , GRSGola significantly improved the prediction (delta C-index [95% confidence interval]: 0.0087 [0.0044, 0.0130]; IDIrel : 0.0509 [0.0131, 0.0894]; cfNRI improved only in cases: 0.1761 [0.0253, 0.3219]). GRSKhera yielded slightly worse prediction results than GRSGola .

Considerations on the Castrop formula for calculation of intraocular lens power

BackgroundTo explain the concept of the Castrop lens power calculation formula and show the application and results from a large dataset compared to classical formulae.MethodsThe Castrop vergence formula is based on a pseudophakic model eye with 4 refractive surfaces. This was compared against the SRKT, Hoffer-Q, Holladay1, simplified Haigis with 1 optimized constant and Haigis formula with 3 optimized constants. A large dataset of preoperative biometric values, lens power data and postoperative refraction data was split into training and test sets. The training data were used for formula constant optimization, and the test data for cross-validation. Constant optimization was performed for all formulae using nonlinear optimization, minimising root mean squared prediction error.ResultsThe constants for all formulae were derived with the Levenberg-Marquardt algorithm. Applying these constants to the test data, the Castrop formula showed a slightly better performance compared to the classical formulae in terms of prediction error and absolute prediction error. Using the Castrop formula, the standard deviation of the prediction error was lowest at 0.45 dpt, and 95% of all eyes in the test data were within the limit of 0.9 dpt of prediction error.ConclusionThe calculation concept of the Castrop formula and one potential option for optimization of the 3 Castrop formula constants (C, H, and R) are presented. In a large dataset of 1452 data points the performance of the Castrop formula was slightly superior to the respective results of the classical formulae such as SRKT, Hoffer-Q, Holladay1 or Haigis.

An Extensive Examination of Discovering 5-Methylcytosine Sites in Genome-Wide DNA Promoters Using Machine Learning Based Approaches

It is well-known that the major reason for the rapid proliferation of cancer cells are the hypomethylation of the whole cancer genome and the hypermethylation of the promoter of particular tumor suppressor genes. Locating 5-methylcytosine (5mC) sites in promoters is therefore a crucial step in further understanding of the relationship between promoter methylation and the regulation of mRNA gene expression. High throughput identification of DNA 5mC in wet lab is still time-consuming and labor-extensive. Thus, finding the 5mC site of genome-wide DNA promoters is still an important task. We compared the effectiveness of the most popular and strong machine learning techniques namely XGBoost, Random Forest, Deep Forest, and Deep Feedforward Neural Network in predicting the 5mC sites of genome-wide DNA promoters. A feature extraction method based on k-mers embeddings learned from a language model were also applied. Overall, the performance of all the surveyed models surpassed deep learning models of the latest studies on the same dataset employing other encoding scheme. Furthermore, the best model achieved AUC scores of 0.962 on both cross-validation and independent test data. We concluded that our approach was efficient for identifying 5mC sites of promoters with high performance.

Identification of colorectal neoplasia by using serum bile acid profile

Background Colonoscopy is the gold standard for detecting earlier stages of CRC, although screening of patients is difficult because of invasiveness, low compliance and procedural health risks. Therefore, the need for new screening methods for CRC is rising. Previous studies have demonstrated the diagnostic ability of serum BAs; however, the results have been inconsistent. In this study, we conducted a comprehensive analysis of serum BAs from patients with CRC and verified their diagnostic ability to detect CRC. Methods A total of 56 CRC patients (n = 14 each of stages I–IV), 59 patients with colonic adenoma and 60 healthy controls were included. Age and sex were matched for each group. Serum BA compositions were measured by LC-MS/MS and serum concentration of 30 types of BAs were analysed by discriminant analysis with multidimensional scaling method. Results Free CA, 3epi-DCA&CDCA, 3-dehydro CA, GCA and TCA were extracted as principal component (PC) 1 and free 3-dehydroDCA as PC 2 by canonical discriminant function coefficients. The verification of discriminability using cross-validation method revealed that the correct classification rate was 66.3% for original data and 52.6% for cross-validation data. Conclusions A combined analysis using comprehensive serum BA concentration can be an efficient method for screening CRC.

Deep Learning for Monitoring Agricultural Drought in South Asia Using Remote Sensing Data

Drought, a climate-related disaster impacting a variety of sectors, poses challenges for millions of people in South Asia. Accurate and complete drought information with a proper monitoring system is very important in revealing the complex nature of drought and its associated factors. In this regard, deep learning is a very promising approach for delineating the non-linear characteristics of drought factors. Therefore, this study aims to monitor drought by employing a deep learning approach with remote sensing data over South Asia from 2001–2016. We considered the precipitation, vegetation, and soil factors for the deep forwarded neural network (DFNN) as model input parameters. The study evaluated agricultural drought using the soil moisture deficit index (SMDI) as a response variable during three crop phenology stages. For a better comparison of deep learning model performance, we adopted two machine learning models, distributed random forest (DRF) and gradient boosting machine (GBM). Results show that the DFNN model outperformed the other two models for SMDI prediction. Furthermore, the results indicated that DFNN captured the drought pattern with high spatial variability across three penology stages. Additionally, the DFNN model showed good stability with its cross-validated data in the training phase, and the estimated SMDI had high correlation coefficient R2 ranges from 0.57~0.90, 0.52~0.94, and 0.49~0.82 during the start of the season (SOS), length of the season (LOS), and end of the season (EOS) respectively. The comparison between inter-annual variability of estimated SMDI and in-situ SPEI (standardized precipitation evapotranspiration index) showed that the estimated SMDI was almost similar to in-situ SPEI. The DFNN model provides comprehensive drought information by producing a consistent spatial distribution of SMDI which establishes the applicability of the DFNN model for drought monitoring.

Under a growth-centered accountability system: A job demand and resource perspective for physical educators

The purpose of this study was to reveal the work life of physical educators who were experiencing a teacher accountability system in the US. A combination of stratified and purposeful sampling procedures was adopted to select a sample of schools that served communities with various socioeconomic backgrounds. The schools were recently incorporated in a state teacher accountability system that used student learning growth for teacher evaluation. A group of 51 certified physical educators was selected from the sampled schools. Data were collected from three sources: an online survey, in-school observations, and semi-structured interviews. The survey and observations cross-validated data related to job description, workload, general working environment, and demographic information. The interviews detailed the teachers’ accounts about their work life, perceptions of the job demands, and resources under the teacher accountability system. Inductive analysis revealed that the lack of critical job resources, especially curricular supports, equipment, and professional development, limited the teachers from promoting student learning that the teacher accountability system required. Teachers were subjected to immense job demands, specifically a large body of content knowledge to teach with shrinking instruction time and ambiguous role. They also felt pressured to incorporate reading, writing, and mathematics in physical education lessons with little or no curricular support. For teachers to teach physical education successfully in this accountability context, systemic changes are needed to provide teachers with adequate professional development and curricular resources to promote learning in physical education.

Development and dosimetric assessment of an automatic dental artifact classification tool to guide artifact management techniques in a fully automated treatment planning workflow.

Purpose:We conducted our study to develop a tool capable of automatically detecting dental artifacts in a CT scan on a slice-by-slice basis and to assess the dosimetric impact of implementing the tool into the Radiation Planning Assistant (RPA), a web-based platform designed to fully automate the radiation therapy treatment planning process.Methods:We developed an automatic dental artifact identification tool and assessed the dosimetric impact of its use in the RPA. Three users manually annotated 83,676 head-and-neck (HN) CT slices (549 patients). Majority-voting was applied to the individual annotations to determine the presence or absence of dental artifacts. The patients were divided into train, cross-validation, and test data sets (ratio: 3:1:1, respectively). A random subset of images without dental artifacts was used to balance classes (1:1) in the training data set. The Inception-V3 deep learning model was trained with the binary cross-entropy loss function. With use of this model, we automatically identified artifacts on 15 RPA HN plans on a slice-by-slice basis and investigated three dental artifact management methods applied before and after volumetric modulated arc therapy (VMAT) plan optimization. The resulting dose distributions and target coverage were quantified.Results:Per-slice accuracy, sensitivity, and specificity were 99 %, 91 %, and 99 %, respectively. The model identified all patients with artifacts. Small dosimetric differences in total plan dose were observed between the various density-override methods (±1 Gy). For the pre- and post-optimized plans, 90 % and 99 %, respectively, of dose comparisons resulted in normal structure dose differences of ±1 Gy. Differences in the volume of structures receiving 95 % of the prescribed dose (V95[%]) were ≤0.25 % for 100 % of plans.Conclusion:The dosimetric impact of applying dental artifact management before and after artifact plan optimization was minor. Our results suggest that not accounting for dental artifacts in the current RPA workflow (where only post-optimization dental artifact management is possible) may result in minor dosimetric differences. If RPA users choose to override CT densities as a solution to managing dental artifacts, our results suggest segmenting the volume of the artifact and overriding its density to water is a safe option.

Classification of Baby Cry Sound Using Higuchi’s Fractal Dimension with K-Nearest Neighbor and Support Vector Machine

The crying baby sound is a way to express their physical and psychological conditions. For most people, it is difficult to distinguish the meaning of the sound of a baby’s cry because they are almost all similar. In fact, sometimes parents still lack knowledge of understanding their baby’s condition from the sound of crying. Therefore, research to classify the types of baby crying sounds mathematically becomes very interesting. This study aimed to classify the types of crying sounds for babies using fractal dimensions, particularly the Higuchi Fractal dimension. The data used in this study were 80 data consisting of 4 types of cries, namely hunger cries, tired cries, stomach ache cries, and uncomfortable cries. In this study k-max values of 10, 16, and 50 were selected as experiments. The results of the fractal dimension value of each sound signal were carried out in the data grouping process. The k-fold cross validation data distribution method with values of k = 5 and 10 were used. The classification process was carried out using the K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) methods. Based on the research results, the best accuracy was 78.75% at level 5 decomposition with 5-fold cross validation, k-max value = 10, and K = 9 value on the KNN method. Whereas with the SVM method, the best accuracy was 80% at level 5 decomposition with 10-fold cross validation, k-max value = 10 and c = 10 and using RBF kernel with γ = 10. So in this study, the Support Vector Machine (SVM) method was slightly better than the K-Nearest Neighbor (KNN) method.

Two-Decade Variability of Sea Surface Temperature and Chlorophyll-a in the Northern South China Sea as Revealed by Reconstructed Cloud-Free Satellite Data

Monthly sea surface temperature (SST) and chlorophyll-a (Chl-a) concentration during 1997–2018 in the Northern South China Sea (NSCS) were gap-filled using the Data INterpolating Empirical Orthogonal Functions (DINEOFs) approach. Good accuracy of the reconstructed data was verified based on an independent cross-validation data set. Significant spatial heterogeneities and seasonal variations were observed in the cloud-free SST and Chl-a data. An apparent decreasing trend in SST was found along the coast of Guangdong, east of Hainan Island, and in the Taiwan Strait with an average decreasing rate amounting to 0.01 °C per year. Increase in SST was observed in Beibu Gulf and the deep offshore region in the NSCS at an average rate as high as 0.009 °C per year. An increasing trend in Chl-a was detected in the northern Beibu Gulf, near the mouth of the Pearl River estuary, and around Shantou with an average rate of 0.005 mg <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\cdot \text{m}^{-3}$ </tex-math></inline-formula> per year. EOF analysis of SST and Chl-a time series showed that the first three modes of both SST and Chl-a can explain most of their variances with accumulative contributions of 99.1% and 82.3%, respectively. The effects of surface net heat flux (NHF), Oceanic Niño Index (ONI), wind speed, and river flux on the first three modes of SST and Chl-a were quantitatively elucidated. Factors that affected the temporal and spatial variations of Chl-a and SST in the NSCS were qualitatively discussed, including El Niño-Southern Oscillation (ENSO), monsoon, upwelling, typhoon, internal waves, and Kuroshio intrusion.

Development of high-resolution future ocean regional projection datasets for coastal applications in Japan

In this study, we developed two high-resolution future ocean regional projection datasets for coastal applications in Japan, in which we made use of dynamical downscaling via regional ocean models with atmospheric forcing from two climate models (i.e., MIROC5 and MRI-CGCM3) participating in Coupled Model Intercomparison Project Phase 5 (CMIP5) under historical, representative concentration pathway (RCP) 2.6, and RCP8.5 scenarios. The first dataset was an eddy-resolving 10-km resolution product covering the North Pacific Ocean area and ranging continuously from 1981 to 2100, in which the Kuroshio current and mesoscale structures were reasonably resolved. The second dataset was a 2-km resolution product covering the regional domain surrounding Japan and comprising 10–15-year time slices, in which the coastal geometry and current structure were resolved even more realistically. An important feature of these datasets was the availability of reference datasets based on atmospheric and oceanic reanalysis data for cross-validation during the historical run period. Using these reference datasets, biases of regional surface thermal properties and the Kuroshio states during the historical run period were evaluated, which constitute important information for users of the datasets. In these downscaled datasets, the future surface thermal responses were generally consistent with those of their original data. Utilizing the high-resolution property of the downscaled data, possible future impact analyses regarding coastal phenomena such as strait throughflows, coastal sea level variability, and the Kuroshio intrusion phenomenon into bays (“Kyucho” phenomenon) were demonstrated and the important role of the Kuroshio state representation was indicated, which had proved difficult to analyze using the low-resolution projection data. Given these properties, the present datasets would be useful in climate change adaptation studies regarding the Japanese coastal region.

Incorporating a transfer learning technique with amino acid embeddings to efficiently predict N-linked glycosylation sites in ion channels

Glycosylation is a dynamic enzymatic process that attaches glycan to proteins or other organic molecules such as lipoproteins. Research has shown that such a process in ion channel proteins plays a fundamental role in modulating ion channel functions. This study used a computational method to predict N-linked glycosylation sites, the most common type, in ion channel proteins. From segments of ion channel proteins centered around N-linked glycosylation sites, the amino acid embedding vectors of each residue were concatenated to create features for prediction. We experimented with two different models for converting amino acids to their corresponding embeddings: one was fed with ion channel sequences and the other with a large dataset composed of more than one million protein sequences. The latter model stemmed from the idea of transfer learning technique and emerged as a more efficient feature extractor. Our best model was obtained from this transfer learning approach and a hyperparameter tuning process with a random search on 5-fold cross-validation data. It achieved an accuracy, specificity, sensitivity, and Matthews correlation coefficient of 93.4%, 92.8%, 98.6%, and 0.726, respectively. Corresponding scores on an independent test were 92.9%, 92.2%, 99%, and 0.717. These results outperform the position-specific scoring matrix features that are predominantly employed in post-translational modification site predictions. Furthermore, compared to N-GlyDE, GlycoEP, SPRINT-Gly, the most recent N-linked glycosylation site predictors, our model yields higher scores on the above 4 metrics, thus further demonstrating the efficiency of our approach.

Automatic Detection of COVID-19 Using a Stacked Denoising Convolutional Autoencoder

The exponential increase in new coronavirus disease 2019 ({COVID-19}) cases and deaths has made COVID-19 the leading cause of death in many countries. Thus, in this study, we propose an efficient technique for the automatic detection of COVID-19 and pneumonia based on X-ray images. A stacked denoising convolutional autoencoder (SDCA) model was proposed to classify X-ray images into three classes: normal, pneumonia, and {COVID-19}. The SDCA model was used to obtain a good representation of the input data and extract the relevant features from noisy images. The proposed model’s architecture mainly composed of eight autoencoders, which were fed to two dense layers and SoftMax classifiers. The proposed model was evaluated with 6356 images from the datasets from different sources. The experiments and evaluation of the proposed model were applied to an 80/20 training/validation split and for five cross-validation data splitting, respectively. The metrics used for the SDCA model were the classification accuracy, precision, sensitivity, and specificity for both schemes. Our results demonstrated the superiority of the proposed model in classifying X-ray images with high accuracy of 96.8%. Therefore, this model can help physicians accelerate COVID-19 diagnosis.