Learning Approach Research Articles

A robust deep learning approach for identification of RNA 5-methyluridine sites

RNA 5-methyluridine (m5U) sites play a significant role in understanding RNA modifications, which influence numerous biological processes such as gene expression and cellular functioning. Consequently, the identification of m5U sites can play a vital role in the integrity, structure, and function of RNA molecules. Therefore, this study introduces GRUpred-m5U, a novel deep learning-based framework based on a gated recurrent unit in mature RNA and full transcript RNA datasets. We used three descriptor groups: nucleic acid composition, pseudo nucleic acid composition, and physicochemical properties, which include five feature extraction methods ENAC, Kmer, DPCP, DPCP type 2, and PseDNC. Initially, we aggregated all the feature extraction methods and created a new merged set. Three hybrid models were developed employing deep-learning methods and evaluated through 10-fold cross-validation with seven evaluation metrics. After a comprehensive evaluation, the GRUpred-m5U model outperformed the other applied models, obtaining 98.41% and 96.70% accuracy on the two datasets, respectively. To our knowledge, the proposed model outperformed all the existing state-of-the-art technology. The proposed supervised machine learning model was evaluated using unsupervised machine learning techniques such as principal component analysis (PCA), and it was observed that the proposed method provided a valid performance for identifying m5U. Considering its multi-layered construction, the GRUpred-m5U model has tremendous potential for future applications in the biological industry. The model, which consisted of neurons processing complicated input, excelled at pattern recognition and produced reliable results. Despite its greater size, the model obtained accurate results, essential in detecting m5U.

A novel classification of dyslipidaemia through the analysis of five million lipid profiles: an unsupervised machine learning approach

Abstract Background Dyslipidemia encompasses a wide range of lipoprotein disorders categorised through two classifications (Fredrickson-Levy [FL] or Sniderman).(1,2) However, both classifications are criticised for relying on incomplete knowledge of lipoprotein metabolism, especially with the emergence of novel treatment options and variations in individual treatment responses.(3) Clustering, an unsupervised machine learning (ML) algorithm that can process a wide range of variables, has the potential to unmask patient groups with distinct molecular profiles and unique therapeutic targets that can inform more effective prevention strategies for cardiovascular disease (CVD).(4) Aim We aimed to use unsupervised ML algorithms to discover intrinsic dyslipidaemia categories from lipoprotein measurements, recognise the necessary components of lipid panels for classification, and analyse the similarities between the newly formed clusters, FL and Sniderman classifications. Methods Lipid profiles of 5,080,248 patients were obtained from the ‘Very Large Database of Lipids’ database. This yielded up to 78 blood components per patient, including at least 31 lipoprotein variables. The analysis involved unsupervised K-means clustering with optimised values for K and the subset of variables, determined in an unsupervised manner using a suitable measure of complexity. We then interpreted our clusters using probabilistic decision trees to provide compact and interpretable representations. Finally, we compared the clusters with Sniderman and FL categories. Results In a completely unsupervised fashion, we identified 14 clusters that could be matched to Sniderman categories. The confusion matrix showed total agreement of 76% (see Figure 1, left panel), relative Cohen’s kappa of 0.78 (the relative version captures accuracy on categories containing smaller numbers of patient profiles) and an accuracy of 96% on the small Type III class. Similar results were observed when matching to FL types. We accurately represented our clusters using probabilistic decision trees of small depth (see Figure 2). We discovered that the data had low intrinsic dimension and a manifold-like structure in which the different clusters could be illustrated (see Figure 1, right panel). Specifically, only 3 variables were needed to obtain our classification: apolipoprotein b, total cholesterol and triglycerides. Conclusion We showed that completely unsupervised ML techniques can uncover dyslipidaemia categories in lipoprotein profiles from a large patient population. The categories largely align with existing classifications based on prior knowledge of lipoprotein metabolism. Furthermore, few lipoprotein variables were required for categorisation (low-dimension data), which could aid in determining which lipoproteins should be measured in a clinical setting. Further analysis of the differences between ML clusters and traditional classifications is needed, which may enhance CVD risk management.

Retraction Note: Brain tumor magnetic resonance image classification: a deep learning approach

Retraction Note: Brain tumor magnetic resonance image classification: a deep learning approach

An Artificial Intelligence approach to monitor infectious diseases: lessons learned from COVID-19

Abstract Background The World Health Organization declared the start (March 11th 2020) and the end (May 5th 2023) of COVID19 pandemic, while in Italy it has been present since February 2020. Indicators such as number of new positive cases, deaths and hospitalizations are used to monitor epidemiological trends, but they suffer from biases limiting their effectiveness. Methods We used data from the Emergency Medical Services Activities as an alternative and studied three types of contributions: COVID19, flu and baseline during three time frames: -period 1 (July 1st-Oct 11th 2016; Feb 10th-May 20th 2017) used to model the baseline, when flu contribution should be negligible. -Period 2: (Dec 15th-Feb 14th 2017), when flu in 2017 is very evident. -Period 3: (March 11th-31st 2020), when the first COVID19 wave is dominant. To extract the pure contribution from flu and COVID19, the baseline contribution was properly subtracted. Results From these data we developed a machine learning approach (MLA) that offers a simple and powerful tool to monitor COVID19 or future infectious diseases. To maximize the identification power of COVID19, we used the Toolkit for multivariate analysis package. An artificial neural network, multilayer perceptron, deep neural network and boosted decision tree methods have been trained, considering the events in period 3 as signal and period 1 as background. To avoid overtraining, the samples were divided in two: one half to train the algorithm, and the other to check the performance. The results are stable over time and able to efficiently discriminate COVID19. With 50% efficiency of accepting COVID19 patients, roughly 95% of baseline patients can be rejected. Conclusions MLA can be used to early assign a probability of COVID19 without specific test, only relying on standard triage and emergency call details. This tool could be very useful to early detect the presence of new pandemics and tag positive patients before the official healthcare reporting system. Key messages • AI model using data from Emergency Medical Services Activities can be a useful tool to early detect new pandemics and tag positive patients before the official healthcare reporting system. • Covid19 pandemic gave us the possibility to develop new digital tools to use in the analysis of epidemiological trends.

Analysis on the Impact of English Teachers' Continuing Professional Development (CPD) on Teaching Ability: A Case Study of Applied Undergraduate Colleges in Henan Province

Background and purpose: With the rapid development of higher education in China and the rise of applied undergraduate colleges, the continuing professional development (CPD) of English teachers is becoming increasingly important. This study aims to explore in depth the impact of CPD on the teaching ability of English teachers in applied undergraduate colleges in Henan Province, analyse teachers' attitudes towards and participation in CPD and its effectiveness, and examine the role of demographic characteristics and teacher creativity in this process. Method: A mixed research method was used, combining quantitative and qualitative analysis. A questionnaire survey was conducted among 294 English teachers (with a 85.5% response rate) from 10 applied undergraduate colleges in Henan Province. The questionnaire included aspects such as teachers' background information, frequency of CPD participation, attitudes towards CPD, and perceived teaching ability. In addition, 20 semi-structured interviews and four focus group discussions (24 participants in total) were conducted to obtain more in-depth qualitative data. Descriptive statistics, correlation analysis and multiple regression analysis were performed using SPSS software, and Hayes' PROCESS macro was used to analyse moderated effects. Main results: Teachers had a generally positive attitude towards CPD (mean score 4.587/5), but institutional support was lower (mean score 3.326/5). There were significant differences in the frequency of participation in CPD activities: professional reading had the highest participation rate (65.9% participated frequently or very frequently), while industry-university cooperation had the lowest (only 22.8%). CPD had a positive impact on all aspects of teaching ability, especially technology integration (72.5% reported a significant or transformative impact). Multiple regression analysis showed that the frequency of CPD participation was the strongest predictor of perceived teaching ability (β = 0.458, p < 0.001), followed by attitude towards CPD (β = 0.265, p < 0.001). Teacher creativity significantly moderated the relationship between CPD participation and teaching competence (B = 0.087, p = 0.026), indicating that teachers with higher creativity are better able to translate their CPD experiences into teaching practice. Heavy workload (65.6%) and insufficient time (59.5%) were the main barriers to CPD participation. Technical training (64.9%) and reflective practice groups (59.2%) were the most popular types of CPD activities. Conclusions and recommendations: The findings of this study highlight the important role of CPD in improving the teaching capacity of English teachers in applied undergraduate institutions, while also revealing the challenges in the current implementation of CPD. Recommendations for educational institutions are as follows: Increase institutional support for CPD to create more time and resources for teachers to participate in CPD. Design a diversified CPD programme with a balanced mix of activities, with a particular focus on industry-university collaboration and action research. Focus on technical training and reflective practice to improve teachers' digital literacy and innovation capabilities. Develop personalised CPD pathways that take into account teachers' career stages, educational backgrounds and career aspirations. Explore more flexible CPD formats, such as microlearning modules or blended learning approaches, to overcome time and workload barriers. This study provides an empirical basis and practical guidance for the CPD of English teachers in applied undergraduate colleges in Henan Province and even in China, which is of great significance for improving teaching quality and student learning outcomes.

Standardization of transthoracic impedance values for estimating heart failure and its utility

Abstract Background The use of intrathoracic impedance for examining pleural effusion in heart failure patients has been explored previously. However, due to concerns about its accuracy and complexity, transthoracic impedance values have had limitations in quantifying extravascular lung water. Our dual objectives were to clarify the relationship between intrathoracic conditions and percutaneous transthoracic impedance values, and to establish a basic model for a new machine learning-based estimation system for assessing intrathoracic conditions in patients with heart failure. Methods First, we developed a live porcine congested lung model that induced pleural effusion by substantial fluid loading, and then longitudinally evaluated its correlation with percutaneous transthoracic impedance values. Second, we conducted multi-frequency bioelectrical impedance analysis to simultaneously collect electrical, physical, and hematological data from 63 hospitalized heart failure patients and 82 healthy volunteers Results In the porcine model, pleural effusion correlated with a concurrent decrease in SpO2 and a gradual decrease in impedance. We indexed and generated features from the measured values and developed an intrathoracic estimation model based on electrical measurements and clinical findings using a decision tree-based machine learning approach. Among the 286 features collected per individual, the model used 16 features. Notably, the model demonstrated high accuracy in discriminating patients with pleural effusion, achieving an AUC of 0.905 (95% CI: 0.870-0.940) in cross-validation, significantly outperforming the conventional frequency-based method. Conclusions Transthoracic impedance values, when indexed, reflect intrathoracic conditions and improve estimation. Our results highlighted the potential of machine learning and thoracic impedance measurement for accurate estimation of pleural effusion. This approach provides an effective means of assessing intrathoracic conditions. Figure: The upper figure shows the relationship between impedance values and pleural effusion in the congested lung model. On these results, human data were collected to construct the estimation model. The lower illustrates the accuracy of the model and provides a comprehensive view of the system.

Predicting Obesity via Deep Learning: The Role of Chrononutrition

Abstract Background Obesity is a significant public health challenge, prompting the need for assessing effective tools for its monitoring and prevention. Chrononutrition investigates the role of timing, regularity and frequency of food intake on body metabolism and health. Here we aim to check if data on the distribution of calorie intake in the day may improve BMI prediction as compared to using only data on total day intake. Methods We apply deep learning on chrononutritional data on calorie intake over 6 daily timeslots from a nationally representative cross-sectional survey sample of the Italian population (INRAN-SCAI 2005/2006) comprising 2313 Italian adults 18-64 ys old with 3 day diet diaries. In particular, we implemented three deep neural network models, varying network configurations and feature selection. The 1st includes 18 (6 for each day diary) chrononutritional variables and the 3 day total intakes, beside age and sex; the 2nd includes age, sex and the 18 chrononutritional variables only; the 3rd uses age, sex and the 3 day totals only. We applied Early Stopping to limit overfitting. Optimization was conducted using the Adam algorithm, minimizing the mean squared error (RMSE), while mean absolute error (MAE) was the primary performance indicator. Results The results highlight that Model 3 outperformed the other models with MAE=2.514 and RMSE=3.476, indicating that for the purpose of predicting obesity, information on the time-of-day distribution of calorie intake is not increasing predictive ability of BMI on top of age, sex and total calorie intake in the day. Conclusions Additional use of day-time intake information worsened prediction of BMI via deep learning using this cross sectional Italian survey. However the error made is still sizeable and warrants the inclusion of other variables; our results don’t rule out the role of chrononutrition on metabolic health which may act through other biological mechanisms related to specific nutrients. Key messages • Chrononutrition data don’t improve BMI prediction over total calorie intake data in a cross-setional representative sample of the Italian population. • Deep learning approach to BMI prediction shows the importance of targeted selection of features in public health nutrition.

Interpreting pretext tasks for active learning: a reinforcement learning approach

As the amount of labeled data increases, the performance of deep neural networks tends to improve. However, annotating a large volume of data can be expensive. Active learning addresses this challenge by selectively annotating unlabeled data. There have been recent attempts to incorporate self-supervised learning into active learning, but there are issues in utilizing the results of self-supervised learning, i.e., it is uncertain how these should be interpreted in the context of active learning. To address this issue, we propose a multi-armed bandit approach to handle the information provided by self-supervised learning in active learning. Furthermore, we devise a data sampling process so that reinforcement learning can be effectively performed. We evaluate the proposed method on various image classification benchmarks, including CIFAR-10, CIFAR-100, Caltech-101, SVHN, and ImageNet, where the proposed method significantly improves previous approaches.

Enhancing Alaskan wildfire prediction and carbon flux estimation: A two-stage deep learning approach within a process-based model

Abstract Wildfires in boreal forests release substantial amounts of carbon into the atmosphere. However, current land-surface models are limited in their representation of fire processes, including their ignition and spread. This study thus developed FireDL, a novel data-driven machine-learning model for the prediction of natural wildfires, and combined it with a land-surface model to better understand the impact of fire on carbon fluxes. FireDL has a two-stage deep learning structure that sequentially combines a long short-term memory (LSTM) algorithm and an artificial neural network (ANN). Preliminary random forest analysis identified fire duration as an important factor in predicting the burned area. Thus, in FireDL, the LSTM algorithm was employed to predict fire occurrence and duration, utilizing lightning, vegetation, and climate datasets. Subsequently, the ANN predicted the total burned area using the LTSM-derived fire duration predictions and climate datasets as input. FireDL produced a robust performance in predicting large fires (>10,000 ha), achieving a correlation coefficient of 0.72. The daily-scaled burned area predictions derived from FireDL were integrated into the Community Land Model version 5 – Biogeochemistry (CLM5-BGC) to produce CLM5-BGC-FireDL. This integration considerably improved carbon emission estimations. Notably, the total net ecosystem exchange (NEE) estimated using CLM5-BGC-FireDL in 2019, the year with the highest recorded burned area during our study, was twice that estimated using the standard CLM5-BGC. Discrepancies in the NEE can significantly influence atmospheric CO2 levels, highlighting the importance of our fire prediction model in forecasting the burned area and carbon emissions. The use of FireDL with future climate scenarios is thus anticipated to yield valuable insights into ecosystem management and climate change mitigation strategies.

School-based outdoor education and teacher well-being: the LärMiljö (Learning Environment) - study

Abstract Background School-based outdoor education (OE) is a teaching method aiming to promote pupils’ learning and wellbeing. OE with its action-oriented and cooperative learning approach can inspire teachers to develop their teaching and may also have a positive impact on their work wellbeing. According to self-determination theory, satisfaction or frustration of employees’ basic psychological needs - need for autonomy, competence and relatedness - are important facilitators or impediments of work wellbeing. We explored possible associations between OE use and teachers’ 1) satisfaction/frustration of basic needs at work and b) work engagement. Methods The study, part of the LärMiljö-study, used data collected in 2022 through electronic surveys among teachers in Swedish-language primary schools in Finland; N = 134 teachers from 44 schools. Main measures were the 24-item Basic Psychological Need Satisfaction and Need Frustration at Work Scale and 9-item Utrecht Work Engagement Scale. Data was analysed with Mann-Whitney U-tests and Spearman correlation. Results The vast majority was female, classroom teachers and had a master’s degree. Preliminary findings show that 69% practiced OE (sporadically or regularly) and 31% did not; both groups were similar in terms of teaching experience, age and gender. Teachers using OE had higher autonomy and competence need satisfaction, and lower relatedness need frustration than teachers not using OE (p < 0.05). There was no difference in overall work engagement (p = 0.07), but OE teachers had higher dedication to work (p < 0.05). Satisfaction of basic needs at work was positively, and frustration of needs negatively, associated with work engagement (p < 0.001). Conclusions Practicing OE appears to have positive associations with teachers’ dedication to work and satisfaction of need for autonomy and competence, and a negative association with relatedness frustration. More research is needed to confirm and examine the direction of these associations. Key messages • Outdoor education (OE) may besides pupils’ learning and wellbeing also promote teachers’ work wellbeing. Work wellbeing among teachers was studied within the framework of self-determination theory. • Practicing OE seems to be positively associated with teachers’ dedication to work, a sub-dimension of work engagement, and satisfaction of the basic needs for autonomy and competence at work.

Automated deep learning approach for identifying etiologies of hypertrophic phenocopies by echocardiography

Abstract Background Left ventricular hypertrophy (LVH) is an indicator of poor prognosis across various cardiovascular diseases and is developed by various etiologies. Transthoracic echocardiography (TTE) is the most available cardiac imaging tool to detect LVH. However, it is still challenging to differentiate three non-ischemic cardiomyopathies which frequently presenting LVH: Fabry cardiomyopathy (FC), hypertrophic cardiomyopathy (HCM), and cardiac amyloidosis (CA) from morphology or cardiac function evaluated by TTE. Purpose We aimed to develop stepwise deep learning models to differentiate FC, HCM, and CA using radiomics extracted from TTE. Methods We conducted a retrospective cohort study between 2011 and 2022 in Taiwan. Patients with left ventricular (LV) mass index≥95g/m² in women or ≥115g/m² in men, or the maximal LV wall thickness≥13mm on echocardiography were enrolled. The TTE automatic processing models included four steps: image quality assessment, view classification, auto-segmentation of interventricular septum (IVS), and dynamic radiomics analysis. One echocardiographic study was scanned per patient, and each study contains multiple cardiac cycles. One sample was extracted from each cycle and all samples were evaluated by image quality assessment and view classification model that based on VGG16. Samples with apical four (A4C) and apical five chamber (A5C) views were retained for following steps. Manual contouring of the septal region served as the ground truth for auto-segmentation based on U-Net. The last step was the extraction of dynamic radiomic features from IVS and the dynamic radiomic features were analyzed using an LSTM model to differentiate three hypertrophic cardiomyopathies. Results There were 103 patients with FC, 96 patients with HCM, and 49 patients with CA in this study. Patients with CA were older and had lower LV ejection fraction (EF) than those with FC or HCM (age: 65.8±9.7 vs. 59.3±11.1 vs. 58.6±14.2 years, P=0.002; LVEF: 56.6±10.8 vs. 61.9±8.8 vs. 64.5±8.3%, P<0.001). In addition, patient with FC had larger LV mass index than those with HCM or CA. (Fig. 2A) The details of the performances of four stepwise models were provided in Figure 2 (B-F). The image quality assessment and view classification models achieved accuracies of 91.7% and 95.2%, respectively. The auto-segmentation had the median Dice score as 0.88 for contouring IVS. The results indicated that deep learning algorithms had great ability to recognize different views with proper image quality, and could contour the IVS adequately. The accuracy of dynamic radiomics-based LSTM model to differentiate three hypertrophic phenocopies was 74.5%. Conclusion We developed an automatic process with four stepwise models for analyzing the radiomics of IVS on TTE to differentiate FC, HCM and CA with the modest performance by LSTM model. The developed automatic models may provide potential implications to identify various etiologies LVH in the future.Figure 1-Workflow of this studyFigure 2-Performances of stepwise models

A Multi-Modal Approach for A Healthier Weight and More Active Lifestyle

Abstract Background Obesity and overweight has been increasing globally and in the UK over the last few years, bringing along a wide array of challenges in health and wellbeing. Not only do they take a heavy toll on those affected, but obesity and overweight-related ill-health is costing the UK National Health Service £6.1 with projected costs to run up to £9.7 billion in 2050. This project aimed to apply artificial intelligence to identify the predictive indicators for obesity and co-create an intervention with public health practitioners and stakeholders. Methods Using health and wellbeing public health population data from the 2015, 2018 and 2022 from a city in the East Midlands, we applied artificial intelligence that uses machine learning techniques to discover insights, find hidden patterns and discover relationships in the data about engaging in physical activity and a healthy diet. The data on dietary choices and physical activity preferences were further analysed using advanced machine learning techniques. Results The outcomes from our machine learning process revealed patterns of physical activity engagement and diet from specific locations, along with a range of demographic variables that influence individuals’ patterns of physical ability and dietary practices. These findings, which will be presented in detail, have been integrated into a comprehensive city map. This map not only showcases opportunities for engaging in a healthy lifestyle but also aids in prioritising interventions in areas of greatest need. Conclusions The AI technology we employed allowed us to parameterise and generalise our findings with the possibility of scaling up the machine learning approach to other similar datasets. The findings facilitated the development of bespoke training for public health officers working with underserved communities, further demonstrating the impact of applying AI to inform the design of public health interventions. Key messages • Applying Artificial Intelligence to public health data has the potential to inform the development of effective interventions and increase precision. • Using AI in public health data can generate new insights into population health and well-being and inform public health policy.

Left atrial function for predicting atrial fibrillation: a machine learning approach

Abstract Background Common risk scores for atrial fibrillation (AF) are based on clinical risk factors, and little is known about the contribution of left atrial (LA) function parameters to risk assessment. Purpose To determine whether a machine learning-based model that includes LA function parameters outperforms a commonly used risk scores, the CHARGE-AF, in predicting AF. Methods Participants in the Atherosclerosis Risk in Communities (ARIC) Study (a community-based cohort study in the USA) who attended visit 5 (2011-2013) and without a history of AF were included in this analysis. Incident AF was ascertained from ECG during study visits, hospitalization discharge codes, and death certificates. Random survival forest methodology was used to build a prediction model for AF; candidate variables included traditional clinical risk factors and measures of LA structure and function. Dataset was randomly split into a training (75%) and testing (25%) cohort. For each variable, importance was measured based on the reduction in Harrell’s c-statistic if the variable was randomly permuted prior to obtaining predicted values. Variables significantly associated with AF were included in the final model. Model calibration was tested using the Greenwood-Nam-D'Agostino (GND) test. Discrimination ability of our model was assessed in the testing cohort using Harrell’s c-statistic and compared with that of the CHARGE-AF model. Results 4,909 participants (59% female, 22% Black, mean age 75 years) were included. During a median follow-up of 7.0 years (1st-3rd quartile: 5.4-7.8), 650 participants developed AF. Variables significantly associated with incident AF were race, triglycerides, systolic blood pressure, coronary heart disease, age, LA maximal volume index, NT-proBNP, and LA reservoir strain (Figure 1). LA reservoir strain was the variable most strongly associated with AF. The final model showed good discrimination (c-statistic (95% CI): 0.734 (0.698-0.768)). The GND test did not detect statistically significant miscalibration (GND p=0.08). Compared with our model, CHARGE-AF had lower discrimination ability (c-statistic (95% CI): 0.635 (0.594-0.677)) and the difference was statistically significant (difference in c-statistic (95% CI): 0.099 (0.043-0.152)). Conclusion A machine learning-based model with few variables including LA function outperforms a commonly used risk model in predicting AF. Incorporation of LA functional assessment may help inform clinicians regarding individual risk of AF.

The Mutscore metapredictor: a new application of artificial intelligence in cardiogenetics

Abstract Introduction The rapid development of high-throughput next generation sequencing has shifted the limits of genetic knowledge from variant sequencing to variant interpretation. The current standards recommend that sequence variants are interpreted according to a comprehensive analysis including, among others, computational data. The two major types of in-silico predictive tools encompass those predicting the theoretical variant effect on splicing and algorithms predicting the potential damage of missense variants. However, overall for missense variants, prediction accuracy is often limited and results are sometimes inconsistent between software programs. Purpose We aimed at testing the predictive performance of a new predictive algorithm (Mutscore) for missense variants based on a machine learning approach, in our cohort of families referred for hereditary cardiac diseases. Methods We retrospectively reviewed DNA sequencing results from 230 families (from 01.07.2014 till 01.07.2023) addressed to our center for channelopathy or hereditary cardiomyopathy. Missense variants were identified and evaluated according to the current standards of the American College of Medical Genetics and Genomics. Variants were analysed with the commonly used in-silico tools CADD, Polyphen, Alpha-missense and Revel. We further applied the Mutscore, a new metapredictor integrating 16 existing predictive tools to data on variant topographical location. Results Among our 230 families, we detected 252 missense variants (class I-II 2.4%, class III 62.3%, class IV and V 15.9% and 19.4% respectively). We observed a significant positive correlation (r = 0.59, p=0.000) between the Mutscore and the interpretation of variants according to standards. The Mutscore had a better predictive performance than Polyphen (AUC 0.83 vs 0.67, p=0.007), Alpha-missense (AUC 0.87 vs 0.79, p=0.047) and CADD (AUC 0.87 vs 0.70, p=0.0001). Compared to Revel, the Mutscore had a comparable predictive performance (0.89 vs 0.87, p=NS), but a better sensitivity (75% vs 66%) at the maximum tolerated false positive rate of 10%. Figure 1. Focusing on variants of uncertain significance (VUS), we applied Mutscore cut-off values which were identified to reclassify variants in our previous study (1). Importantly, the Mutscore reclassified 45% of VUS into likely benign/pathogenic variants. Figure 2. Conclusions The Mutscore, through its metaprediction approach integrating data on variant topographical location, improves the accuracy of pathogenicity prediction as compared to three out of four algorithms commonly used in clinical practice, and seems to represent a valuable tool contributing to VUS disambiguation.

Metabolic modelling as a powerful tool to identify critical components of Pneumocystis growth medium.

Establishing suitable in vitro culture conditions for microorganisms is crucial for dissecting their biology and empowering potential applications. However, a significant number of bacterial and fungal species, including Pneumocystis jirovecii, remain unculturable, hampering research efforts. P. jirovecii is a deadly pathogen of humans that causes life-threatening pneumonia in immunocompromised individuals and transplant patients. Despite the major impact of Pneumocystis on human health, limited progress has been made in dissecting the pathobiology of this fungus. This is largely due to the fact that its experimental dissection has been constrained by the inability to culture the organism in vitro. We present a comprehensive in silico genome-scale metabolic model of Pneumocystis growth and metabolism, to identify metabolic requirements and imbalances that hinder growth in vitro. We utilise recently published genome data and available information in the literature as well as bioinformatics and software tools to develop and validate the model. In addition, we employ relaxed Flux Balance Analysis and Reinforcement Learning approaches to make predictions regarding metabolic fluxes and to identify critical components of the Pneumocystis growth medium. Our findings offer insights into the biology of Pneumocystis and provide a novel strategy to overcome the longstanding challenge of culturing this pathogen in vitro.

Concertorl: A reinforcement learning approach for finite-time single-life enhanced control and its application to direct-drive tandem-wing experiment platforms

Concertorl: A reinforcement learning approach for finite-time single-life enhanced control and its application to direct-drive tandem-wing experiment platforms

An adaptive physics-informed deep learning approach for structural nonlinear response prediction

An adaptive physics-informed deep learning approach for structural nonlinear response prediction

Predicting the success of startups using a machine learning approach

Successful investment in early-stage companies has high uncertainty. More specifically, the tools available to investors need to be more robust to reduce the risk and manage the uncertainty of startups. This research aims to use machine learning methods to design a prediction solution to identify successful startups for investors. In order to design the predicting solution and provide policies, classification, and clustering algorithms have been utilized to predict the success of startups and perform feature importance analysis based on the SHAP and permutation methods. Subsequently, the performance of four classification algorithms, such as Random Forest, Gradient Boost, Multilayer Perceptron, Logistic Regression and Support Vector Machine, are compared to predict business success. Meanwhile, Random Forest and Gradient Boosting algorithms showed the best accuracy, which was equal to 82% and 80%, respectively. Based on the feature importance of Random Forest and Gradient Boosting, which is obtained from the SHAP method, indicated that the higher values of “Number of followers on LinkedIn”, “Number of employees on LinkedIn”, “Number of followers on Twitter”, and “Last raised amount” have higher SHAP values and a more significant impact on the model output. Three clustering algorithms including hierarchy, K-means, and DBSCAN were also compared. Among them, the K-means algorithm performs best with 72% silhouette, and K-means was employed to explain each cluster’s characteristics. Finally, an effective artificial intelligence-based prediction solution has been proposed to show the way for investors to apply machine learning concepts to predict the success of startups.

A Dataset of 3M Single-DOF Planar 4-, 6-, and 8-bar Linkage Mechanisms with Open and Closed Coupler Curves for Machine Learning-Driven Path Synthesis

Abstract In recent years, there has been a strong interest in applying machine learning techniques to path synthesis of linkage mechanisms. However, progress has been stymied due to a scarcity of high-quality datasets. In this paper, we present a comprehensive dataset comprising nearly three million samples of four-, six-, and eight-bar linkage mechanisms with open and closed coupler curves. Current machine learning approaches to path synthesis also lack standardized metrics for evaluating outcomes. To address this gap, we propose six key metrics to quantify results, providing a foundational framework for researchers to compare new models with existing ones. We also present a Variational AutoEncoder based model in conjunction with a k-nearest neighbor search approach to demonstrate the utility of our dataset. In the end, we provide example mechanisms that generate various curves along with numerical evaluation of the proposed metrics.

Deep learning modeling of RNA ac4C deposition reveals the importance of plant alternative splicing.

The N4-acetylcytidine (ac4C) modification has recently been characterized as a noncanonical RNA marker in plants. While the precise installation of ac4C sites in individual plant transcripts continues to present challenges, the biological roles of ac4C in specific plant species are gradually being deciphered. Herein, we utilized a deep learning technique called iac4C (intelligent ac4C) to predict ac4C sites in mRNA. ac4C deposition was effectively forecasted by the iac4C model (AUROC = 0.948), revealing a reliable distribution pattern primarily situated in the transcribing area as opposed to regions that are not translated. The iac4C deep learning approach using a combination of BiGRU and self-attention mechanisms both validates previous studies showing a positive correlation between ac4C and RNA splicing in plant species and reveals new examples of other splicing events associated with ac4C. Our advanced deep learning algorithm for analyzing ac4C enables swift identification of important biological phenomena that would otherwise be challenging to uncover through traditional experimental approaches. These findings provide insight into the essential regulatory function of site-specific ac4C deposition in alternative splicing processes. The source code and datasets for iac4C are available at https://github.com/xlwei507/iac4C .