Computational Learning Methods Research Articles

Editorial: Computational modeling and machine learning methods in neurodevelopment and neurodegeneration: from basic research to clinical applications

Editorial: Computational modeling and machine learning methods in neurodevelopment and neurodegeneration: from basic research to clinical applications

Neoantigen prioritization based on antigen processing and presentation

Somatic mutations in tumor cells give rise to mutant proteins, fragments of which are often presented by MHC and serve as neoantigens. Neoantigens are tumor-specific and not expressed in healthy tissues, making them attractive targets for T-cell-based cancer immunotherapy. On the other hand, since most somatic mutations differ from patient to patient, neoantigen-targeted immunotherapy is personalized medicine and requires their identification in each patient. Computational algorithms and machine learning methods have been developed to prioritize neoantigen candidates. In fact, since the number of clinically relevant neoantigens present in a patient is generally limited, this process is like finding a needle in a haystack. Nevertheless, MHC presentation of neoantigens is not random but follows certain rules, and the efficiency of neoantigen detection may be further improved with technological innovations. In this review, we discuss current approaches to the detection of clinically relevant neoantigens, with a focus on antigen processing and presentation.

Computational Advances in Ionic Liquid Applications for Green Chemistry: A Critical Review of Lignin Processing and Machine Learning Approaches.

The valorization and dissolution of lignin using ionic liquids (ILs) is critical for developing sustainable biorefineries and a circular bioeconomy. This review aims to critically assess the current state of computational and machine learning methods for understanding and optimizing IL-based lignin dissolution and valorization processes reported since 2022. The paper examines various computational approaches, from quantum chemistry to machine learning, highlighting their strengths, limitations, and recent advances in predicting and optimizing lignin-IL interactions. Key themes include the challenges in accurately modeling lignin's complex structure, the development of efficient screening methodologies for ionic liquids to enhance lignin dissolution and valorization processes, and the integration of machine learning with quantum calculations. These computational advances will drive progress in IL-based lignin valorization by providing deeper molecular-level insights and facilitating the rapid screening of novel IL-lignin systems.

Learning meaningful representation of single-neuron morphology via large-scale pre-training.

Single-neuron morphology, the study of the structure, form, and shape of a group of specialized cells in the nervous system, is of vital importance to define the type of neurons, assess changes in neuronal development and aging and determine the effects of brain disorders and treatments. Despite the recent surge in the amount of available neuron morphology reconstructions due to advancements in microscopy imaging, existing computational and deep learning methods for modeling neuron morphology have been limited in both scale and accuracy. In this paper, we propose MorphRep, a model for learning meaningful representation of neuron morphology pre-trained with over 250 000 existing neuron morphology data. By encoding the neuron morphology into graph-structured data, using graph transformers for feature encoding and enforcing the consistency between multiple augmented views of neuron morphology, MorphRep achieves the state of the art performance on widely used benchmarking datasets. Meanwhile, MorphRep can accurately characterize the neuron morphology space across neuron morphometrics, fine-grained cell types, brain regions and ages. Furthermore, MorphRep can be applied to distinguish neurons under a wide range of conditions, including genetic perturbation, drug injection, environment change and disease. In summary, MorphRep provides an effective strategy to embed and represent neuron morphology and can be a valuable tool in integrating cell morphology into single-cell multiomics analysis. The codebase has been deposited in https://github.com/YaxuanLi-cn/MorphRep.

Assessing interpretability of data‐driven fuzzy models: Application in industrial regression problems

AbstractMachine Learning (ML) has attracted great interest in the modeling of systems using computational learning methods, being utilized in a wide range of advanced fields due to its ability and efficiency to process large amounts of data and to make predictions or decisions with a high degree of accuracy. However, with the increase in the complexity of the models, ML's methods have presented complex structures that are not always transparent to the users. In this sense, it is important to study how to counteract this trend and explore ways to increase the interpretability of these models, precisely where decision‐making plays a central role. This work addresses this challenge by assessing the interpretability and explainability of fuzzy‐based models. The structural and semantic factors that impact the interpretability of fuzzy systems are examined. Various metrics have been studied to address this topic, such as the Co‐firing Based Comprehensibility Index (COFCI), Nauck Index, Similarity Index, and Membership Function Center Index. These metrics were assessed across different datasets on three fuzzy‐based models: (i) a model designed with Fuzzy c‐Means and Least Squares Method, (ii) Adaptive‐Network‐based Fuzzy Inference System (ANFIS), and (iii) Generalized Additive Model Zero‐Order Takagi‐Sugeno (GAM‐ZOTS). The study conducted in this work culminates in a new comprehensive interpretability metric that covers different domains associated with interpretability in fuzzy‐based models. When addressing interpretability, one of the challenges lies in balancing high accuracy with interpretability, as these two goals often conflict. In this context, experimental evaluations were performed in many scenarios using 4 datasets varying the model parameters in order to find a compromise between interpretability and accuracy.

Convolutional Neural Network Processing of Radio Emission for Nuclear Composition Classification of Ultra-High-Energy Cosmic Rays

Ultra-high-energy cosmic rays (UHECRs) are extremely rare energetic particles of ordinary matter in the Universe, traveling astronomical distances before reaching the Earth’s atmosphere. When primary cosmic rays interact with atmospheric nuclei, cascading extensive air showers (EASs) of secondary elementary particles are developed. Radio detectors have proven to be a reliable method for reconstructing the properties of EASs, such as the shower’s axis, its energy, and its maximum (Xmax). This aids in understanding fundamental astrophysical phenomena, like active galactic nuclei and gamma-ray bursts. Concurrently, data science has become indispensable in UHECR research. By applying statistical, computational, and deep learning methods to both real-world and simulated radio data, researchers can extract insights and make predictions. We introduce a convolutional neural network (CNN) architecture designed to classify simulated air shower events as either being generated by protons or by iron nuclei. The classification achieved a stable test error of 10%, with Accuracy and F1 scores of 0.9 and an MCC of 0.8. These metrics indicate strong prediction capability for UHECR’s nuclear composition, based on data that can be gathered by detectors at the world’s largest cosmic rays experiment on Earth, the Pierre Auger Observatory, which includes radio antennas, water Cherenkov detectors, and fluorescence telescopes.

Automatic Content Analysis Systems: Detecting Disinformation in Social Networks

In the 21st century, the rapid rise of disinformation and propaganda has become a significant global issue, undermining democratic processes and socio-political institutions. Disinformation, defined as intentionally false or misleading information, aims to manipulate public opinion and cause economic harm. This paper explores the use of computational linguistics and machine learning methods to detect disinformation. Techniques such as text preprocessing, feature extraction, and classification algorithms (e.g., SVM, naive Bayes) are adapted for identifying fake news. Recent studies demonstrate the effectiveness of these methods in social media and news platforms, highlighting the importance of advanced models like GPT-4 in improving detection accuracy and combating the spread of disinformation.

FEP Augmentation as a Means to Solve Data Paucity Problems for Machine Learning in Chemical Biology.

In the realm of medicinal chemistry, the primary objective is to swiftly optimize a multitude of chemical properties of a set of compounds to yield a clinical candidate poised for clinical trials. In recent years, two computational techniques, machine learning (ML) and physics-based methods, have evolved substantially and are now frequently incorporated into the medicinal chemist's toolbox to enhance the efficiency of both hit optimization and candidate design. Both computational methods come with their own set of limitations, and they are often used independently of each other. ML's capability to screen extensive compound libraries expediently is tempered by its reliance on quality data, which can be scarce especially during early-stage optimization. Contrarily, physics-based approaches like free energy perturbation (FEP) are frequently constrained by low throughput and high cost by comparison; however, physics-based methods are capable of making highly accurate binding affinity predictions. In this study, we harnessed the strength of FEP to overcome data paucity in ML by generating virtual activity data sets which then inform the training of algorithms. Here, we show that ML algorithms trained with an FEP-augmented data set could achieve comparable predictive accuracy to data sets trained on experimental data from biological assays. Throughout the paper, we emphasize key mechanistic considerations that must be taken into account when aiming to augment data sets and lay the groundwork for successful implementation. Ultimately, the study advocates for the synergy of physics-based methods and ML to expedite the lead optimization process. We believe that the physics-based augmentation of ML will significantly benefit drug discovery, as these techniques continue to evolve.

Machine learning model based on RCA-PDCA nursing methods and differentiating factors to predict hypotension during cesarean section surgery

BackgroundIntraoperative hypotension during cesarean section has become a serious complication for maternal and fetal healthy. It is commonly encountered by subarachnoid anesthesia. However, currently used control methods have varying degrees of side effects, such as drugs. The Root Cause Analysis (RCA) - Plan, Do, Check, Act (PDCA) is a new model of care that identifies the root causes of problems. The study aimed to demonstrate the usefulness of RCA-PDCA nursing methods in preventing intraoperative hypotension during cesarean section and to predict the occurrence of intraoperative hypotension through a machine learning model. MethodsPatients who underwent cesarean section at Traditional Chinese Medicine of Southwest Medical University from January 2023 to December 2023 were retrospectively screened, and the data of their gestational times, age, height, weight, history of allergies, intraoperative vital signs, fetal condition, operative time, fluid out and in, adverse effects, use of vasopressor drugs, anxiety-depression-pain scores, and satisfaction were collected and analyzed. The statistically different features were screened and five machine learning models were used as predictive models to assess the usefulness of the RCA-PDCA model of care. Results(1) Compared with the general nursing model, the RCA-PDCA nursing model significantly reduces the incidence of intraoperative hypotension and postoperative complications in cesarean delivery, and the patient experience is comfortable and satisfactory. (2) Among the five machine learning models, the RF model has the best predictive performance, and the accuracy of the random forest model in preventing intraoperative hypotension is as high as 90%. ConclusionThrough computer machine learning model analysis, we prove the importance of the RCA-PDCA nursing method in the prevention of intraoperative hypotension during cesarean section, especially the Random Forest model which performed well and promoted the application of artificial intelligence computer learning methods in the field of medical analysis.

Experimental, Computational, and Machine Learning Methods for Prediction of Residual Stresses in Laser Additive Manufacturing: A Critical Review.

In recent decades, laser additive manufacturing has seen rapid development and has been applied to various fields, including the aerospace, automotive, and biomedical industries. However, the residual stresses that form during the manufacturing process can lead to defects in the printed parts, such as distortion and cracking. Therefore, accurately predicting residual stresses is crucial for preventing part failure and ensuring product quality. This critical review covers the fundamental aspects and formation mechanisms of residual stresses. It also extensively discusses the prediction of residual stresses utilizing experimental, computational, and machine learning methods. Finally, the review addresses the challenges and future directions in predicting residual stresses in laser additive manufacturing.

Hate speech detection in social media: Techniques, recent trends, and future challenges

AbstractThe realm of Natural Language Processing and Text Mining has seen a surge in interest from researchers in hate speech detection, leading to an increase in related studies. This analysis aims to create a valuable resource by summarizing the methods and strategies used to combat hate speech in social media. We perform a detailed review to achieve a deep knowledge of the hate speech detection landscape from 2018 to 2023, revealing global incidents of hate speech in 2022–2023. Sixty‐six relevant articles were selected for this review. Existing studies were analyzed and categorized into five method categories: Machine Learning, Deep Learning, Ensemble models, Graph Neural Networks, and Graph Convolutional Networks. These advancements can aid social networking services in identifying hate messages before being posted, reducing the risk of harassment. The review also covers available hate speech datasets and highlights research challenges, but it is clear that a definitive solution to this problem is yet to be found. Future research directions are recommended to address the ongoing challenges in Hate Speech Detection.This article is categorized under: Applications of Computational Statistics > Computational Linguistics Statistical Learning and Exploratory Methods of the Data Sciences > Knowledge Discovery Statistical Learning and Exploratory Methods of the Data Sciences > Classification and Regression Trees (CART) Statistical Learning and Exploratory Methods of the Data Sciences > Text Mining

Grain and grain boundary segmentation using machine learning with real and generated datasets

We report a significantly improved accuracy in grain boundary segmentation using Convolutional Neural Networks (CNN) trained on a combination of real and generated data. Manual segmentation is accurate but time-consuming, and existing computational methods are faster but often inaccurate. To combat this dilemma, machine learning models can be used to achieve the accuracy of manual segmentation and have the efficiency of a computational method. An extensive dataset of from 316L stainless steel samples is additively manufactured, prepared, polished, etched, and then microstructure grain images were systematically collected. Grain segmentation via existing computational methods and manual (by-hand) were conducted, to create ”real” training data. A Voronoi tessellation pattern combined with random synthetic noise and simulated defects, is developed to create a novel artificial grain image fabrication method. This provided training data supplementation for data-intensive machine learning methods. The accuracy of the grain measurements from microstructure images segmented via computational methods and machine learning methods proposed in this work are calculated and compared, and also provide benchmarks in grain segmentation. Over 400 images of the microstructure of stainless steel samples were manually segmented for machine learning training applications. This data and the artificial data is available on Kaggle.

Avaliação de recursos para o mapeamento de cobertura do solo em sub-bacia do rio Capibaribe-PE usando imagem Kompsat-2

Land use and land cover mapping is an important factor in geospatial analysis in watershed management. The integration of remote sensing images and Machine Learning classification techniques enable the identification and environmental monitoring of landscape elements. The MSC (MultiSpectral Camara) sensor on the Kompsat-2 satellite captures images of high spatial resolution, which allows the identification of terrestrial resources on a local scale. Six data models were developed for classifying land use and land cover by Random Forest in a Capibaribe River sub-basin. These models were created based on spectral indices and ranking of variable importance. The evaluation of the results was done through spatial quantification and accuracy analysis. Products based on bands and spectral indices showed global accuracy ranging between 94 and 98%, where the classes of Arboreal and Shrubby Vegetation stood out with estimates of accuracy of the producer and user above 80%. Products with the lowest data resources showed poor accuracy performance with overall accuracy values clustered below 60%. This study is the first to use adaptations of Kompsat-2 spectral data and computer learning methods to demonstrate the application of high-performance land cover mapping. Thus, this article contributed to the monitoring of the soil surface in urban sub-basins that need precise spatial information about the state of environmental conversation.

A hybrid study of a 4-stage compressed solar distiller based on experimental, computational and deep learning methods

In the present study, a hybrid technique is employed to demonstrate the thermal performance of a 4-stage compressed solar distiller with hydrophilic evaporators (CSDHE) regarding the role of natural convection in the air gap domain. Initially, the test is conducted under simulated solar radiation of 2.13 W·m−2 over 5400 s. A thermodynamic model incorporating diffusion is developed for the next step to analyze internal transient temperatures. In the third step, a 3D single-phase natural convection model of the CSDHE is simulated to achieve the research objectives. Based on experimental, thermal modeling, and computational results, a dual deep neural network is developed to predict the thermal behavior of the system and the variation of convection heat transfer coefficient for different air gap thicknesses over 10,000 s. The experimental results revealed a distilled water production rate of 3.806 kg·m−2h−1 with a gained output ratio (GOR) of 112 %. Moreover, the findings demonstrate that conduction accounts for 84.23 % of total heat transfer, while convection accounts for 6.36 % during 5400 s. We also found that, by employing a convolutional neural network (CNN) and directly harvesting data from pre-defined contours, the hybrid training dataset for deep forward neural network (DFNN) is trained 28.35 % faster than conventional methods.

Improving the classification of cardinality phenotypes using collections

MotivationPhenotypes are observable characteristics of an organism and they can be highly variable. Information about phenotypes is collected in a clinical context to characterize disease, and is also collected in model organisms and stored in model organism databases where they are used to understand gene functions. Phenotype data is also used in computational data analysis and machine learning methods to provide novel insights into disease mechanisms and support personalized diagnosis of disease. For mammalian organisms and in a clinical context, ontologies such as the Human Phenotype Ontology and the Mammalian Phenotype Ontology are widely used to formally and precisely describe phenotypes. We specifically analyze axioms pertaining to phenotypes of collections of entities within a body, and we find that some of the axioms in phenotype ontologies lead to inferences that may not accurately reflect the underlying biological phenomena.ResultsWe reformulate the phenotypes of collections of entities using an ontological theory of collections. By reformulating phenotypes of collections in phenotypes ontologies, we avoid potentially incorrect inferences pertaining to the cardinality of these collections. We apply our method to two phenotype ontologies and show that the reformulation not only removes some problematic inferences but also quantitatively improves biological data analysis.

Exploring the influence of nasal vestibule structure on nasal obstruction using CFD and Machine Learning method

Motivated by clinical findings about the nasal vestibule, this study analyzes the aerodynamic characteristics of the nasal vestibule and attempt to determine anatomical features which have a large influence on airflow through a combination of Computational Fluid Dynamics (CFD) and machine learning method. Firstly, the aerodynamic characteristics of the nasal vestibule are detailedly analyzed using the CFD method. Based on CFD simulation results, we divide the nasal vestibule into two types with distinctly different airflow patterns, which is consistent with clinical findings. Secondly, we explore the relationship between anatomical features and aerodynamic characteristics by developing a novel machine learning model which could predict airflow patterns based on several anatomical features. Feature mining is performed to determine the anatomical feature which has the greatest impact on respiratory function. The method is developed and validated on 41 unilateral nasal vestibules from 26 patients with nasal obstruction. The correctness of the CFD analysis and the developed model is verified by comparing them with clinical findings.

Basing on the machine learning model to analyse the coronary calcification score and the coronary flow reserve score to evaluate the degree of coronary artery stenosis

AimTo obtain the coronary artery calcium score (CACS) for each branch in coronary artery computed tomography angiography (CCTA) examination combined with the flow fraction reserve (FFR) of each branch in the coronary artery detected by CT and apply a machine learning model (ML) to analyse and predict the severity of coronary artery stenosis. MethodsAll patients who underwent coronary computed tomography angiography (CCTA) from January 2019 to April 2022 in the HOSPITAL (T.C.M) AFFILIATED TO SOUTHWEST MEDICAL UNIVERSITY) were retrospectively screened, and their sex, age, characteristics of lipid-containing lesions, coronary calcium score (CACS) and CT-FFR values were collected. Five machine learning models, random forest (RF), k-nearest neighbour algorithm (KNN), kernel logistic regression, support vector machine (SVM) and radial basis function neural network (RBFNN), were used as predictive models to evaluate the severity of coronary stenosis. ResultsAmong the five machine learning models, the SVM model achieved the best prediction performance, and the prediction accuracy of mild stenosis was up to 90%. Second, age and male sex were important influencing factors of increasing CACS and decreasing CT-FFR. Moreover, the critical CACS value of myocardial ischemia >200.70 was calculated. ConclusionThrough computer machine learning model analysis, we prove the importance of CACS and FFR in predicting coronary stenosis, especially the prominent vector machine model, which promotes the application of artificial intelligence computer learning methods in the field of medical analysis.

A Need For Speed: Enhancing F1 Race Cars with a Novel Computational Fluid Dynamics and Machine Learning Method

Formula One (F1) engineers dedicate their lives to finding tenths of a second in car performance. But, the technical complexities of F1 cars must now also make inroads to increasing sustainability because fuel consumption in F1 is a major problem. Furthermore, F1 teams with lower budgets now need to find ways to be more efficient in aerodynamic testing to match the richer teams’ expensive equipment. This would improve the competitiveness of F1 and please 100 million fans around the world. Computer-aided design to model Formula One race cars for aerodynamic testing, computational fluid dynamics to test 90 different airfoils and simulation setups as well as the finalized model car against the control car, and machine learning to quickly predict the results of other simulations I have not conducted yet with high accuracy rates in a very little amount of time.

Structural optimization of multistage centrifugal pump via computational fluid dynamics and machine learning method

Abstract To implement energy savings in multistage centrifugal pumps, a return channel is utilized to replace the origin inter-stage flow channel structure, and then a single-objective optimization work containing high-precision numerical simulation, design variable dimensionality reduction, and machine learning is conducted to obtain the optimal geometric parameters. The variable dimensionality reduction process is based on the Spearman correlation analysis method. The influence of 15 design variables of the impeller and return channel is investigated, and seven of them with high-impact factors are selected as the final optimization variables. Thereafter, a genetic algorithm-backpropagation neural network (GA-BPNN) model is used to create a surrogate model with a high-fitting performance by employing a GA to optimize the initial thresholds and weights of a BPNN. Finally, a multi-island genetic algorithm (MIGA) is employed to maximize hydraulic efficiency under the nominal condition. The findings demonstrate that the optimized model’s efficiency is increased by 4.29% at 1.0Qd, and the deterioration of the pump performance under overload conditions is effectively eliminated (the maximum efficiency increase is 14.72% at 1.3Qd). Furthermore, the internal flow analysis indicates that the optimization scheme can improve the turbulence kinetic energy distribution and reduce unstable flow structures in the multistage centrifugal pump.

An Expedited Route to Optical and Electronic Properties at Finite Temperature via Unsupervised Learning.

Electronic properties and absorption spectra are the grounds to investigate molecular electronic states and their interactions with the environment. Modeling and computations are required for the molecular understanding and design strategies of photo-active materials and sensors. However, the interpretation of such properties demands expensive computations and dealing with the interplay of electronic excited states with the conformational freedom of the chromophores in complex matrices (i.e., solvents, biomolecules, crystals) at finite temperature. Computational protocols combining time dependent density functional theory and ab initio molecular dynamics (MD) have become very powerful in this field, although they require still a large number of computations for a detailed reproduction of electronic properties, such as band shapes. Besides the ongoing research in more traditional computational chemistry fields, data analysis and machine learning methods have been increasingly employed as complementary approaches for efficient data exploration, prediction and model development, starting from the data resulting from MD simulations and electronic structure calculations. In this work, dataset reduction capabilities by unsupervised clustering techniques applied to MD trajectories are proposed and tested for the ab initio modeling of electronic absorption spectra of two challenging case studies: a non-covalent charge-transfer dimer and a ruthenium complex in solution at room temperature. The K-medoids clustering technique is applied and is proven to be able to reduce by ∼100 times the total cost of excited state calculations on an MD sampling with no loss in the accuracy and it also provides an easier understanding of the representative structures (medoids) to be analyzed on the molecular scale.