Artificial intelligence in cryo-EM: emerging deep neural network methods from sample preparation, particle picking, map reconstruction, modelling to enhanced resolution

Summary A meticulous interpretation of steady-state or unsteady-state relative permeability (Kr) experimental data is required to determine a complete set of Kr curves. In this work, different machine learning (ML) models were developed to assist in a faster estimation of these curves from steady-state drainage coreflooding experimental runs. These ML algorithms include gradient boosting (GB), random forest (RF), extreme gradient boosting (XGB), and deep neural network (DNN) with a main focus on and comparison of the two latter algorithms (XGB and DNN). Based on existing mathematical models, a leading-edge framework was developed where a large database of Kr and capillary pressure (Pc) curves were generated. This database was used to perform thousands of coreflood simulation runs representing oil-water drainage steady-state experiments. The results obtained from these simulation runs, mainly pressure drop along with other conventional core analysis data, were used to estimate analytical Kr curves based on Darcy’s law. These analytically estimated Kr curves along with the previously generated Pc curves were fed as features into the ML model. The entire data set was split into 80% for training and 20% for testing. The k-fold cross-validation technique was applied to increase the model’s accuracy by splitting 80% of the training data into 10 folds. In this manner, for each of the 10 experiments, nine folds were used for training and the remaining fold was used for model validation. Once the model was trained and validated, it was subjected to blind testing on the remaining 20% of the data set. The ML model learns to capture fluid flow behavior inside the core from the training data set. In terms of applicability of these ML models, two sets of experimental data were needed as input; the first was the analytically estimated Kr curves from the steady-state drainage coreflooding experiments, while the other was the Pc curves estimated from centrifuge or mercury injection capillary pressure (MICP) measurements. The trained/tested model was then able to estimate Kr curves based on the experimental results fed as input. Furthermore, to test the performance of the ML model when only one set of experimental data is available to an end user, a recurrent neural network (RNN) algorithm was trained/tested to predict Kr curves in the absence of Pc curves as an input. The performance of the three developed models (XGB, DNN, and RNN) was assessed using the values of the coefficient of determination (R2) along with the loss calculated during training/validation of the model. The respective crossplots along with comparisons of ground truth vs. artificial intelligence (AI)-predicted curves indicated that the model is capable of making accurate predictions with an error percentage between 0.2% and 0.6% on history-matching experimental data for all three tested ML techniques. This implies that the AI-based model exhibits better efficiency and reliability in determining Kr curves when compared to conventional methods. The developed ML models by no means replace the need to conduct drainage coreflooding or centrifuge experiments but act as an alternative to existing commercial platforms that are used to interpret experimental data to predict Kr curves. The two main advantages of the developed ML models are their capability of predicting Kr curves within a matter of a few minutes as well as with limited intervention from the end user. The results also include a comparison between classical ML approaches, shallow neural networks, and DNNs in terms of accuracy in predicting the final Kr curves. The research presented here is an extension of the state-of-the-art framework proposed by Mathew et al. (2021). However, the two main aspects of the current study are the application of deep learning for the prediction of Kr curves and the application of feature engineering. The latter not only reduces the training/testing time for the ML models but also enables the end user to obtain the final predictions with the least set of experimental data. The various models discussed in this research work currently focus on the prediction of Kr curves for drainage steady-state experiments; however, the work can be extended to capture the imbibition cycle as well.

Cryo-electron microscopy (cryo-EM) and cryo-electron tomography (cryo-ET) have transformed structural biology by enabling the visualization of biological macromolecules at near-atomic resolution. Since its establishment in 2002, the Electron Microscopy Data Bank (EMDB) has played a pivotal role in this “Resolution Revolution,” serving as a central repository for cryo-EM data. Over the past decade, major advances in detector technology, image processing, and analytical methodologies have led to a surge in cryo-EM structure depositions. This study provides a comprehensive analysis of temporal trends in the EMDB, highlighting the growing number of entries, improvements in resolution, and the evolution of detector usage. It emphasizes the rise of single-particle analysis (SPA) as the dominant cryo-EM technique and examines how high-resolution detectors have enhanced data quality. Additionally, it explores resolution standards across leading scientific journals, underscoring the role of EMDB not only as a data archive but also as a catalyst for innovation in structural biology.

Role of artificial intelligence in advancing cryo electron microscopy.

Aim: The majority of machine learning (ML) models in healthcare are built on retrospective data, much of which is collected without explicit patient consent for use in artificial intelligence (AI) and ML applications. The primary aim of this study was to evaluate whether clinicians and scientific researchers themselves consent to provide their own data for the training of ML models. Materials and Methods: This survey was conducted through an anonymous online survey, utilizing platforms such as Telegram, LinkedIn, and Viber. The target audience comprised specific international groups, primarily Russian, German, and English-speaking, of clinicians and scientific researchers. These participants ranged in their levels of expertise and experience, from beginners to veterans. The survey centered on a singular, pivotal question: “Do You Consent to the Use of Your Biological and Private Data for Training Machine Learning and AI Models?” Respondents had the option to choose from three responses: “Yes” and “No”. Results: The survey was conducted in January 2024. A total of 119 unique and verified individuals participated in the survey. The results revealed that only 50% of respondents (63 persons) expressed consent to provide their own data for the training of ML and AI models. Conclusion: In the development of ML and AI models, particularly open-source ones, it is crucial to ascertain whether participants are willing to provide their private data. While ML algorithms can transform the nature of data, it is important to remember that the primary owner of this data is the individual. Our findings show that in 50% of the cases, even participants from scientific research and clinical backgrounds – individuals typically accountable for ensuring data quality in AI and ML model development – do not consent to the use of their data in AI and ML settings. This highlights the need for more stringent consent processes and ethical considerations in the utilization of personal data in AI and ML research.

Artificial Intelligence (AI) in healthcare enables predicting diabetes using data-driven methods instead of the traditional ways of screening the disease, which include hemoglobin A1c (HbA1c), oral glucose tolerance test (OGTT), and fasting plasma glucose (FPG) screening techniques, which are invasive and limited in scale. Machine learning (ML) and deep neural network (DNN) models that use large datasets to learn the complex, nonlinear feature interactions, but the conventional ML algorithms are data sensitive and often show unstable predictive accuracy. Conversely, DNN models are more robust, though the ability to reach a high accuracy rate consistently on heterogeneous datasets is still an open challenge. For predicting diabetes, this work proposed a hybrid DNN approach by integrating a bidirectional long short-term memory (BiLSTM) network with a bidirectional gated recurrent unit (BiGRU). A robust DL model, developed by combining various datasets with weighted coefficients, dense operations in the connection of deep layers, and the output aggregation using batch normalization and dropout functions to avoid overfitting. The goal of this hybrid model is better generalization and consistency among various datasets, which facilitates the effective management and early intervention. The proposed DNN model exhibits an excellent predictive performance as compared to the state-of-the-art and baseline ML and DNN models for diabetes prediction tasks. The robust performance indicates the possible usefulness of DL-based models in the development of disease prediction in healthcare and other areas that demand high-quality analytics.

Animal-origin fibers exhibit characteristics that determine their quality, price, and comfort, with diameter and medullation type being particularly important. Medullated fibers are considered problematic for the textile industry as they increase the sensation of itchiness on the skin and the heterogeneity of fabrics, leading to a decrease in fiber quality. Thus, having practical instruments to identify them is very essential. On the other hand, some fibers, depending on the species they come from, are considerably cheaper than others. Hence the existence of counterfeit products, which are detrimental to legitimate producers and consumers alike. Fourier-Transform Infrared (FTIR) spectroscopy can extract a characteristic waveform from fibers than can be later used for classification. However, visually inspecting such waveforms is imprecise. In parallel, Artificial Intelligence is an emerging field that could be useful in this area, offering new possibilities and significant advances. Therefore, the first aim of this thesis was to develop and validate a novel automatic system (referred to as S-Fiber Med) for medullation and diameter assessment of animal fibers based on artificial intelligence. The second objective was to develop and validate two machine learning models, namely, Deep Neural Networks (DNN) and Support Vector Machine (SVM) to classify spectra of fibers by species. For the first objective, the medullation of 88 samples of alpaca, llama and mohair fibers (41, 43 and 4, respectively) was evaluated. Additionally, 269 samples of alpacas were considered for average fiber diameter (AFD) and the results were compared with the Portable Fiber Tester (PFT) and Optical Fibre Diameter Analyser (OFDA) methods (72 and 197 samples, respectively). For the second objective, fiber spectra were acquired using an FTIR spectrometer in Attenuated Total Reflectance (ATR) mode (FTIR-ATR). Camelid (alpaca: n = 51, llama: n=50, vicuña: n = 50) and goat (mohair: n = 35 and cashmere: n = 20) samples were evaluated, from which 1,236 FTIR-ATR spectra were obtained. Results of our first objective indicate that there was no significant difference between medullation percentage obtained with the projection microscope and the S-Fiber Med. The Pearson and Concordance correlation analysis shows a strong, high and significant relationship (p-value < 0.001). The AFDs of alpaca and llama fiber samples obtained with the two methods are very similar, because no significant difference was found at the t-test (p-value>0.172), and they have a strong, high and significant relationship between them, given the high Pearson correlation value (r ≥ 0.96 with p-value <0.001), high Concordance coefficient and bias correction factor. Similar results were found when PFT and OFDA100 were compared with S-Fiber Med. While in the results of our second objective, some visual differences were observed between the spectra of the different species. Accuracies up to 96.75% and 95.12% were obtained when evaluating the DNN and SVM models. Furthermore, an accuracy of 97.8% was obtained when evaluating the FTIR-ATR spectra of South American Camelids (SAC) fibers with DNN, and 97.2% when evaluating them with SVM. A 100% accuracy was obtained when evaluating the FTIR-ATR spectra of vicuña fibers with both models. No significant differences were found (p-value=0.368) by comparing the number of hits against the total number of alpaca, llama, vicuña, mohair and cashmere fibers using DNN. As per the results, it seems that DNN is more accurate than SVM. In conclusion, a novel system was developed that provides rapid (40 seconds/sample), precise, accurate, and automatic measurements of medullation and AFD. Additionally, FTIR-ATR spectrometry techniques combined with machine learning models are a reliable alternative for the identification of SAC and goats through the spectrum of their fibers.

Optimisation and interpretation of machine and deep learning models for improved water quality management in Lake Loktak

Structural studies of biocomplexes using single-particle cryo-electron microscopy (cryo-EM) is now a well-established technique in structural biology and has become competitive with X-ray crystallography. The latest advances in EM enable us to determine structures of protein complexes at 3-5 Å resolution for an extremely broad range of sizes from ~200 kDa up to hundreds of megadaltons (Bartesaghi et al., Science 348(6239):1147-1151, 2051; Bai et al., Nature 525(7568):212-217, 2015; Vinothkumar et al., Nature 515(7525):80-84, 2014; Grigorieff and Harrison, Curr Opin Struct Biol 21(2):265-273, 2011). The majority of biocomplexes comprise a number of different components and are not amenable to crystallisation. Secretion systems are typical examples of such multi-protein complexes, and structural studies of them are extremely challenging. The only feasible approach to revealing their spatial organisation and functional modification is cryo-EM. The development of systems for digital registration of images and algorithms for the fast and efficient processing of recorded images and subsequent analysis facilitated the determination of structures at near-atomic resolution. In this review we will describe sample preparation for cryo-EM, how data are collected by new detectors, and the logistics of image analysis through the basic steps required for reconstructions of both small and large biological complexes and their refinement to nearly atomic resolution. The processing workflow is illustrated using examples of EM analysis of a Type IV Secretion System.

Artificial General Intelligence is the idea that someday an hypothetical agent will arise from artificial intelligence (AI) progresses, and will surpass by far the brightest and most gifted human minds. This idea has been around since the early development of AI. Since then, scenarios on how such AI may behave towards humans have been the subject of many fictional and research works. This paper analyzes the current state of artificial intelligence progresses, and how the current AI race with the ever faster release of impressive new AI methods (that can deceive humans, outperform them at tasks we thought impossible to tackle by AI a mere decade ago, and that disrupt the job market) have raised concerns that Artificial General Intelligence (AGI) might be coming faster that we thought. In particular, we focus on 3 specific families of modern AIs to develop the idea that deep neural networks, which are the current backbone of nearly all artificial intelligence methods, are poor candidates for any AGI to arise due to their many limitations, and therefore that any threat coming from the recent AI race does not lie in AGI but in the limitations, uses, and lack of regulations of our current models and algorithms. This article appears in the AI & Society track.

A novel approach for water quality classification based on the integration of deep learning and feature extraction techniques

The advent of the fourth industrial revolution (industry 4.0) has sparked a transformation in manufacturing through the integration of deep learning (DL) and big data technologies. These innovations have enhanced the ability of manufacturers to process and analyze vast amount of data, providing insights that improve decision-making, efficiency, and overall productivity. Deep learning, a subset of artificial intelligence (AI), offers advanced capabilities in data pattern recognition and predictive modeling, while big data facilitates the management of large and complex datasets from various sources. However, literature on the integration of deep learning, big data and industry 4.0 is still limited in the manufacturing context. This paper provides a detailed overview of big data, deep learning and industry 4.0 in manufacturing. It explores the applications of deep learning and big data in manufacturing, highlighting their role in optimizing production processes, predictive maintenance, quality control, and supply chain management. Furthermore, the paper addresses the key challenges and solutions associated with the integration of these technologies, such as data privacy, security, computational complexity, and the need for skilled labor. KEYWORDS: Deep learning, Industry 4.0, Supply chain management, Big data, Predictive maintenance, Artificial intelligence, ERP systems.

In this study, machine learning (ML), a subdivision of artificial intelligence (AI), is implemented to study the mechanical behavior of adhesive single-lap joints (SLJs) subjected to tensile loading. The experimental data for training and testing the ML models are compiled from peer-reviewed journal papers to eliminate bias and increase the diversity within the data. The dataset is comprised of eight continuous SLJ parameters, which are used to predict the SLJ damage mode and failure strength. To accomplish this, regression and classification models are built using deep neural networks (DNN) and random forests (RF). Finite element (FE) modeling is conducted, and the performance is compared with the accuracy of the regression ML models. Results show ML models were able to predict strength with higher accuracy than FE modeling. Furthermore, both DNN and RF classified damage mode accurately without failure criteria, exposing limitations within FE modeling. As a result, this study introduces the use of ML for strength and damage mode prediction of adhesive SLJ, providing insights into their mechanical behavior, revealing hidden property performance patterns, and enhancing predictability.

Machine learning algorithms for deeper understanding and better design of composite adhesive joints

HADA: An automated tool for hardware dimensioning of AI applications

Artificial Intelligence and machine learning (ML) methods are promising for risk-stratification, but the added benefit over traditional statistical methods remains unclear. We compared predictive models developed using machine learning (ML) methods to the Canadian Syncope Risk Score (CSRS), a risk-tool developed with logistic regression for predicting serious adverse events (SAE) after emergency department (ED) disposition for syncope. We used the prospective multicenter cohort data collected for CSRS development at 11 Canadian EDs over an 8-year period to develop four ML models to predict 30-day SAE (death, arrhythmias, MI, structural heart disease, pulmonary embolism, hemorrhage) after ED disposition. The CSRS derivation and validation cohorts were used for training and testing, respectively, and the 43 variables used included demographics, medical history, vital signs, ECG findings, blood tests and the diagnostic impression of the emergency physician. Performance was assessed using the area under the receiver-operating-characteristics curve (AUC) and calibration curves. Of the 4030 patients in the training set and 3819 patients in the test set overall, 286 (3.6%) patients suffered 30-day SAE. The AUCs for model validation in test data were CSRS 0.902 (0.877-0.926), regularized regression 0.903 (0.877-0.928), gradient boosting 0.914 (0.894-0.934), deep neural network 0.906 (0.883-0.929), simplified gradient boosting 0.904 (0.881-0.927). The AUCs and calibration slopes for the ML models and CSRS were similar. Two ML models used fewer predictors than the CSRS but matched its performance. Overall,the ML models matched the CSRS in performance, with some models using fewer predictors.