Search

Background: Mutation processes leave different signatures in genes. For single-base substitutions, previous studies have suggested that mutation signatures are not only reflected in mutation bases but also in neighboring bases. However, because of the lack of a method to identify features of long sequences next to mutation bases, the understanding of how flanking sequences influence mutation signatures is limited.Methods: We constructed a long short-term memory – self organizing map (LSTM-SOM) unsupervised neural network. By extracting mutated sequence features via LSTM and clustering similar features with the SOM, single-base substitutions in The Cancer Genome Atlas database were clustered according to both their mutation site and flanking sequences. The relationship between mutation sequence signatures and clinical features was then analyzed. Finally, we clustered patients into different classes according to the composition of the mutation sequence signatures by the K-means method and then studied the differences in clinical features and survival between classes.Results: Ten classes of mutant sequence signatures (mutation blots, MBs) were obtained from 2,141,527 single-base substitutions via LSTM-SOM machine learning approach. Different features in mutation bases and flanking sequences were revealed among MBs. MBs reflect both the site and pathological features of cancers. MBs were related to clinical features, including age, gender, and cancer stage. The class of an MB in a given gene was associated with survival. Finally, patients were clustered into 7 classes according to the MB composition. Significant differences in survival and clinical features were observed among different patient classes.Conclusions: We provided a method for analyzing the characteristics of mutant sequences. Result of this study showed that flanking sequences, together with mutation bases, shape the signatures of SBSs. MBs were shown related to clinical features and survival of cancer patients. Composition of MBs is a feasible predictive factor of clinical prognosis. Further study of the mechanism of MBs related to cancer characteristics is suggested.

Objectives: This study describes an unsupervised machine learning approach used to estimate the homeostatic model assessment-insulin resistance (HOMA-IR) cut-off for identifying subjects at risk of IR in a given ethnic group based on the clinical data of a representative sample. Methods: The approach was applied to analyse the clinical data of individuals with Arab ancestors, which was obtained from a family study conducted in Nizwa, Oman, between January 2000 and December 2004. First, HOMA-IR-correlated variables were identified to which a clustering algorithm was applied. Two clusters having the smallest overlap in their HOMA-IR values were retrieved. These clusters represented the samples of two populations, which are insulin-sensitive subjects and individuals at risk of IR. The cut-off value was estimated from intersections of the Gaussian functions, thereby modelling the HOMA-IR distributions of these populations. Results: A HOMA-IR cut-off value of 1.62 ± 0.06 was identified. The validity of this cut-off was demonstrated by showing the following: 1) that the clinical characteristics of the identified groups matched the published research findings regarding IR; 2) that a strong relationship exists between the segmentations resulting from the proposed cut-off and those resulting from the two-hour glucose cut-off recommended by the World Health Organization for detecting prediabetes. Finally, the method was also able to identify the cut-off values for similar problems (e.g. fasting sugar cut-off for prediabetes). Conclusion: The proposed method defines a HOMA-IR cut-off value for detecting individuals at risk of IR. Such methods can identify high-risk individuals at an early stage, which may prevent or delay the onset of chronic diseases such as type 2 diabetes. Keywords: Unsupervised Machine Learning; Cluster Analysis; Insulin Resistance; Diabetes Mellitus, Type II.

Previous studies on enhancers and their target genes were largely based on bulk samples that represent ‘average’ regulatory activities from a large population of millions of cells, masking the heterogeneity and important effects from the sub-populations. In recent years, single-cell sequencing technology has enabled the profiling of open chromatin accessibility at the single-cell level (scATAC-seq), which can be used to annotate the enhancers and promoters in specific cell types. A comprehensive resource is highly desirable for exploring how the enhancers regulate the target genes at the single-cell level. Hence, we designed a single-cell database scEnhancer (http://enhanceratlas.net/scenhancer/), covering 14 527 776 enhancers and 63 658 600 enhancer-gene interactions from 1 196 906 single cells across 775 tissue/cell types in three species. An unsupervised learning method was employed to sort and combine tens or hundreds of single cells in each tissue/cell type to obtain the consensus enhancers. In addition, we utilized a cis-regulatory network algorithm to identify the enhancer-gene connections. Finally, we provided a user-friendly platform with seven useful modules to search, visualize, and browse the enhancers/genes. This database will facilitate the research community towards a functional analysis of enhancers at the single-cell level.

Clustering tumor metastasis samples from gene expression data at the whole genome level remains an arduous challenge, in particular, when the number of experimental samples is small and the number of genes is huge. We focus on the prediction of the epithelial-mesenchymal transition (EMT), which is an underlying mechanism of tumor metastasis, here, rather than tumor metastasis itself, to avoid confounding effects of uncertainties derived from various factors. In this paper, we propose a novel model in predicting EMT based on multidimensional scaling (MDS) strategies and integrating entropy and random matrix detection strategies to determine the optimal reduced number of dimension in low dimensional space. We verified our proposed model with the gene expression data for EMT samples of breast cancer and the experimental results demonstrated the superiority over state-of-the-art clustering methods. Furthermore, we developed a novel feature extraction method for selecting the significant genes and predicting the tumor metastasis. The source code is available at "https://github.com/yushanqiu/yushan.qiu-szu.edu.cn".

Background: Identifying clusters of physical activity (PA) from accelerometer data is important to identify levels of sedentary behaviour and physical activity associated with risks of serious health conditions and time spent engaging in healthy PA. Unsupervised machine learning models can capture PA in everyday free-living activity without the need for labelled data. However, there is scant research addressing the selection of features from accelerometer data. The aim of this systematic review is to summarise feature selection techniques applied in studies concerned with unsupervised machine learning of accelerometer-based device obtained physical activity, and to identify commonly used features identified through these techniques. Feature selection methods can reduce the complexity and computational burden of these models by removing less important features and assist in understanding the relative importance of feature sets and individual features in clustering. Method: We conducted a systematic search of Pubmed, Medline, Google Scholar, Scopus, Arxiv and Web of Science databases to identify studies published before January 2021 which used feature selection methods to derive PA clusters using unsupervised machine learning models. Results: A total of 13 studies were eligible for inclusion within the review. The most popular feature selection techniques were Principal Component Analysis (PCA) and correlation-based methods, with k-means frequently used in clustering accelerometer data. Cluster quality evaluation methods were diverse, including both external(e.g. cluster purity) or internal evaluation measures (silhouette score most frequently). Only four of the 13studies had more than 25 participants and only four studies included two or more datasets. Conclusion: There is a need to assess multiple feature selection methods upon large cohort data consisting of multiple (3 or more) PA datasets. The cut-off criteria e.g. number of components, pairwise correlation value, explained variance ratio for PCA, etc. should be expressly stated along with any hyperparameters used in clustering. Refereed/Peer-reviewed

Purpose The deformable registration of 3D chest computed tomography (CT) images is one of the most important tasks in the field of medical image registration. However, the nonlinear deformation and large‐scale displacement of lung tissues caused by respiratory motion cause great challenges in the deformable registration of 3D lung CT images. Materials and methods We proposed an end‐to‐end fast registration method based on unsupervised learning, optimized the classic U‐Net, and added inception modules between skip connections. The inception module attempts to capture and merge information at different spatial scales to generate a high‐precision dense displacement vector field. To solve the problem of voxel folding in flexible registration, we put the Jacobian regularization term into the loss function to directly penalize the singularity of the displacement field during training to ensure a smooth displacement vector field. In the stage of data preprocessing, we segmented the lung fields to eliminate the interference of irrelevant information in the network during training. The existing publicly available datasets cannot implement model training. To alleviate the problem of overfitting caused by limited data resources being available, we proposed a data augmentation method based on the 3D‐TPS (3D thin plate spline) transform to expand the training data. Results Compared with the experimental results obtained by using the VoxelMorph deep learning method and registration packages, such as ANTs and Elastix, we achieved a competitive target registration error of 2.09 mm, an optimal Dice score of 0.987, and almost no folding voxels. Additionally, the proposed method was much faster than the traditional methods. Conclusions In this study, we have shown that the proposed method was efficient in 3D chest image registration. The promising results demonstrated that our method showed strong robustness in the deformable registration of 3D chest CT images.

Purpose To study and investigate the synergistic benefit of incorporating both conventional handcrafted and learning-based features in disease identification across a wide range of clinical setups. Methods and materials In this retrospective study, we collected 170, 150, 209, and 137 patients with four different disease types associated with identification objectives : Lymph node metastasis status of gastric cancer (GC), 5-year survival status of patients with high-grade osteosarcoma (HOS), early recurrence status of intrahepatic cholangiocarcinoma (ICC), and pathological grades of pancreatic neuroendocrine tumors (pNETs). Computed tomography (CT) and magnetic resonance imaging (MRI) were used to derive image features for GC/HOS/pNETs and ICC, respectively. In each study, 67 universal handcrafted features and study-specific features based on the sparse autoencoder (SAE) method were extracted and fed into the subsequent feature selection and learning model to predict the corresponding disease identification. Models using handcrafted alone, SAE alone, and hybrid features were optimized and their performance was compared. Prominent features were analyzed both qualitatively and quantitatively to generate study-specific and cross-study insight. In addition to direct performance gain assessment, correlation analysis was performed to assess the complementarity between handcrafted features and SAE features. Results On the independent hold-off test, the handcrafted, SAE, and hybrid features based prediction yielded area under the curve of 0.761 versus 0.769 versus 0.829 for GC, 0.629 versus 0.740 versus 0.709 for HOS, 0.717 versus 0.718 versus 0.758 for ICC, and 0.739 versus 0.715 versus 0.771 for pNETs studies, respectively. In three out of the four studies, prediction using the hybrid features yields the best performance, demonstrating the general benefit in using hybrid features. Prediction with SAE features alone had the best performance in the HOS study, which may be explained by the complexity of HOS prognosis and the possibility of a slight overfit due to higher correlation between handcrafted and SAE features. Conclusion This study demonstrated the general benefit of combing handcrafted and learning-based features in radiomics modeling. It also clearly illustrates the task-specific and data-specific dependency on the performance gain and suggests that while the common methodology of feature combination may be applied across various studies and tasks, study-specific feature selection and model optimization are still necessary to achieve high accuracy and robustness.

Deep learning architectures are an extremely powerful tool for recognizing and classifying images. However, they require supervised learning and normally work on vectors of the size of image pixels and produce the best results when trained on millions of object images. To help mitigate these issues, we propose an end-to-end architecture that fuses bottom-up saliency and top-down attention with an object recognition module to focus on relevant data and learn important features that can later be fine-tuned for a specific task, employing only unsupervised learning. In addition, by utilizing a virtual fovea that focuses on relevant portions of the data, the training speed can be greatly improved. We test the performance of the proposed Gamma saliency technique on the Toronto and CAT 2000 databases, and the foveated vision in the large Street View House Numbers (SVHN) database. The results with foveated vision show that Gamma saliency performs at the same level as the best alternative algorithms while being computationally faster. The results in SVHN show that our unsupervised cognitive architecture is comparable to fully supervised methods and that saliency also improves CNN performance if desired. Finally, we develop and test a top-down attention mechanism based on the Gamma saliency applied to the top layer of CNNs to facilitate scene understanding in multi-object cluttered images. We show that the extra information from top-down saliency is capable of speeding up the extraction of digits in the cluttered multidigit MNIST data set, corroborating the important role of top down attention.

Objective This study aimed to identify sleep disturbance subtypes (“phenotypes”) among Latinx adults based on objective sleep data using a flexible unsupervised machine learning technique. Methods This study was an analysis of sleep data from three cross-sectional studies of the Precision in Symptom Self-Management Center at Columbia University. All studies focused on sleep health in Latinx adults at increased risk for sleep disturbance. Data on total sleep time (TST), time in bed (TIB), wake after sleep onset (WASO), sleep efficiency (SE), number of awakenings (NOA) and the mean length of nightly awakenings were collected using wrist-mounted accelerometers. Cluster analysis of the sleep data was conducted using an unsupervised machine learning approach that relies on mixtures of multivariate generalized linear mixed models. Results The analytic sample included 494 days of data from 118 adults (Ages 19–77). A 3-cluster model provided the best fit based on deviance indices (ie, DΔ∼ −75 and −17 from 1- and 2- to 3-cluster models, respectively) and likelihood ratio (Pdiff ∼ 0.93). Phenotype 1 (n = 64) was associated with greater likelihood of overall adequate SE and less variability in SE and WASO. Phenotype 2 (n = 11) was characterized by higher NOAs, and greater WASO and TIB than the other phenotypes. Phenotype 3 (n = 43) was characterized by greater variability in SE, bed times and awakening times. Conclusion Robust digital data-driven modeling approaches can be useful for detecting sleep phenotypes from heterogenous patient populations, and have implications for designing precision sleep health strategies for management and early detection of sleep problems.

ObjectivesIt is important to determine an accurate demarcation line (DL) between the cancerous lesions and background mucosa in magnifying narrow-band imaging (M-NBI)-based diagnosis. However, it is difficult for novice endoscopists. Our aim was to automatically determine the accurate DL using a machine learning method.MethodsWe used an unsupervised machine learning approach to determine the DLs because it can reduce the burden of training machine learning models and labeling large datasets. Our method consists of the following four steps: 1) An M-NBI image is segmented into superpixels (a group of neighboring pixels) using simple linear iterative clustering. 2) The image features are extracted for each superpixel. 3) The superpixels are grouped into several clusters using the k-means method. 4) The boundaries of the clusters are extracted as DL candidates. To validate the proposed method, 23 M-NBI images of 11 cases were used for performance evaluation. The evaluation investigated the similarity of the DLs identified by endoscopists and our method, and the Euclidean distance between the two DLs was calculated. For the single case of 11 cases, the histopathological examination was also conducted and was used to evaluate the proposed system.ResultsThe average Euclidean distances for the 11 cases were10.65, 11.97, 7.82, 8.46, 8.59, 9.72, 12.20, 9.06, 22.86, 8.45, and 25.36. The results indicated that the specific selection of the number of clusters enabled the proposed method to detect DLs that were similar to those of the endoscopists. The DLs identified by our method represented the complex shapes of the DLs, similarly to those identified by experienced doctors. Also, it was confirmed that the proposed system could generate the pathologically valid DLs by increasing the number of clusters.ConclusionsOur proposed system can support the training of inexperienced doctors, as well as enrich the knowledge of experienced doctors in endoscopy.

BACKGROUND: Growing use of social media has led to the emergence of virtual controlled substance and illicit drug marketplaces, prompting calls for action by government and law enforcement. Previous studies have analyzed Instagram drug selling via posts. However, comments made by users involving potential drug selling have not been analyzed. In this study, we use unsupervised machine learning to detect and classify prescription and illicit drug-related buying and selling interactions on Instagram. METHODS: We used over 1,000 drug-related hashtags on Instagram to collect a total of 43,607 Instagram comments between February 1(st), 2019 and May 31st, 2019 using data mining approaches in the Python programming language. We then used an unsupervised machine learning approach, the Biterm Topic Model (BTM), to thematically summarize Instagram comments into distinct topic groupings, which were then extracted and manually annotated to detect buying and selling comments. RESULTS: We detected 5,589 comments from sellers, prospective buyers, and online pharmacies from 531 unique posts. The vast majority (99.7%) of comments originated from drug sellers and online pharmacies. Key themes from comments included providing contact information through encrypted third-party messaging platforms, drug availability, and price inquiry. Commonly offered drugs for sale included scheduled controlled substances such as Adderall and Xanax, as well as illicit hallucinogens and stimulants. Comments from prospective buyers of drugs most commonly included inquiries about price and availability. CONCLUSIONS: We detected prescription controlled substances and other illicit drug selling interactions via Instagram comments to posts. We observed that comments were primarily used by sellers offering drugs, and typically not by prospective buyers interacting with sellers. Further research is needed to characterize these “social” drug marketplace interactions on this and other popular social media platforms.

Despite fluorescent cell-labelling being widely employed in biomedical studies, some of its drawbacks are inevitable, with unsuitable fluorescent probes or probes inducing a functional change being the main limitations. Consequently, the demand for and development of label-free methodologies to classify cells is strong and its impact on precision medicine is relevant. Towards this end, high-throughput techniques for cell mechanical phenotyping have been proposed to get a multidimensional biophysical characterization of single cells. With this motivation, our goal here is to investigate the extent to which an unsupervised machine learning methodology, which is applied exclusively on morpho-rheological markers obtained by real-time deformability and fluorescence cytometry (RT-FDC), can address the difficult task of providing label-free discrimination of reticulocytes from mature red blood cells. We focused on this problem, since the characterization of reticulocytes (their percentage and cellular features) in the blood is vital in multiple human disease conditions, especially bone-marrow disorders such as anemia and leukemia. Our approach reports promising label-free results in the classification of reticulocytes from mature red blood cells, and it represents a step forward in the development of high-throughput morpho-rheological-based methodologies for the computational categorization of single cells. Besides, our methodology can be an alternative but also a complementary method to integrate with existing cell-labelling techniques., 13 pages, 3 figures, 4 tables

The aim of this study was to examine the risk factors for pharyngeal carriage of meningococci in third-level students using an unsupervised machine learning approach. Data were gathered as part of meningococcal prevalence studies conducted by the Irish Meningitis and Sepsis Reference Laboratory (IMSRL). Pharyngeal swab cultures for meningococcal carriage were taken from each student once they had completed a single-page anonymous questionnaire addressing basic demographics, social behaviors, living arrangements, vaccination, and antibiotic history. Data were analyzed using multiple correspondence analysis through a machine learning approach. In total, 16,285 students who had a pharyngeal throat swab taken returned a fully completed questionnaire. Overall, meningococcal carriage rate was 20.6%, and the carriage of MenW was 1.9% (n = 323). Young Irish adults aged under 20 years and immunized with the meningococcal C vaccine had a higher MenW colonization rate (n = 171/1260, 13.5%) compared with non-Irish adults aged 20 years or older without the MenC vaccine (n = 5/81, 6%, chi-square = 3.6, p = .05). Unsupervised machine learning provides a useful technique to explore meningococcal carriage risk factors. The issue is very complex, and asked risk factors only explain a small proportion of the carriage. This technique could be used on other conditions to explore reasons for carriage.

Machine learning (ML) is unlocking patterns and insight into data to provide financial value and knowledge for organizations. Use of machine learning in manufacturing environments is increasing, yet sometimes these applications fail to produce meaningful results. A critical review of how defects are classified is needed to appropriately apply machine learning in a production foundry and other manufacturing processes. Four elements associated with defect classification are proposed: Binary Acceptance Specifications, Stochastic Formation of Defects, Secondary Process Variation, and Visual Defect Inspection. These four elements create data space overlap, which influences the bias associated with training supervised machine learning algorithms. If this influence is significant enough, the predicted error of the model exceeds a critical error threshold (CET). There is no financial motivation to implement the ML model in the manufacturing environment if its error is greater than the CET. The goal is to bring awareness to these four elements, define the critical error threshold, and offer guidance and future study recommendations on data collection and machine learning that will increase the success of ML within manufacturing.

Background Pericoronary adipose tissue attenuation expressed by fat attenuation index (FAI) on coronary CT angiography (CCTA) reflects pericoronary inflammation and is associated with cardiac mortality. Objective The aim of this study was to define the sub-phenotypes of coronary CCTA-defined plaque and whole vessel quantification by unsupervised machine learning (ML) and its prognostic impact when combined with pericoronary inflammation. Methods A total of 220 left anterior descending arteries (LAD) with intermediate stenosis who underwent fractional flow reserve (FFR) measurement and CCTA were studied. After removal of outcome and FAI data, the phenotype heterogeneity of CCTA-defined plaque and whole vessel quantification was investigated by unsupervised hierarchical clustering analysis based on Ward's method. Detailed features of CCTA findings were assessed according to the clusters (CS1 and CS2). Major adverse cardiac events (MACE)-free survivals were assessed according to the stratifications by FAI and the clusters. Results Compared with CS2 (n = 119), CS1 (n = 101) were characterized by greater vessel size, increased plaque volume, and high-risk plaque features. FAI was significantly higher in CS1. ROC analyses revealed that best cut-off value of FAI to predict MACE was −73.1. Kaplan-Meier analysis revealed that lesions with FAI ≥ -73.1 had a significantly higher risk of MACE. Multivariate Cox proportional hazards regression analysis revealed that age, FAI ≥ -73.1, and the clusters were independent predictors of MACE. Conclusion Unsupervised hierarchical clustering analysis revealed two distinct CCTA-defined subgroups and discriminated by high-risk plaque features and increased FAI. The risk of MACE differs significantly according to the increased FAI and ML-defined clusters.

Background: Alzheimer’s Disease (AD) is an irreversible, progressive brain disorder that slowly destroys memory and thinking skills. The ability to correctly predict the diagnosis of Alzheimer’s disease in its earliest stages can help physicians make more informed clinical decisions on therapy plans. Objective: This study aimed to determine whether the unsupervised discovering of latent classes of subjects with Mild Cognitive Impairment (MCI) may be useful in finding different prodromal AD stages and/or subjects with a low MCI to AD conversion risk. Methods: Total 18 features relevant to the MCI to AD conversion process led to the identification of 681 subjects with early MCI. Subjects were divided into training (70%) and validation (30%) sets. Subjects from the training set were analyzed using consensus clustering, and Gaussian Mixture Models (GMM) were used to describe the latent classes. The discovered GMM predicted the latent class of the validation set. Finally, descriptive statistics, rates of conversion, and Odds Ratios (OR) were computed for each discovered class. Results: Through consensus clustering, we discovered three different clusters among MCI subjects. The three clusters were associated with low-risk (OR = 0.12, 95%CI = 0.04 to 0.3|), medium-risk (OR = 1.33, 95%CI = 0.75 to 2.37), and high-risk (OR = 3.02, 95%CI = 1.64 to 5.57) of converting from MCI to AD, with the high-risk and low-risk groups highly contrasting. Hence, prodromal AD subjects were present in only two clusters. Conclusion: We successfully discovered three different latent classes among MCI subjects with varied risks of MCI-to-AD conversion through consensus clustering. Two of the discovered classes may represent two different prodromal presentations of Alzheimer´s disease.

Background Drug–target interaction (DTI) plays a vital role in drug discovery. Identifying drug–target interactions related to wet-lab experiments are costly, laborious, and time-consuming. Therefore, computational methods to predict drug–target interactions are an essential task in the drug discovery process. Meanwhile, computational methods can reduce search space by proposing potential drugs already validated on wet-lab experiments. Recently, deep learning-based methods in drug-target interaction prediction have gotten more attention. Traditionally, DTI prediction methods' performance heavily depends on additional information, such as protein sequence and molecular structure of the drug, as well as deep supervised learning. Results This paper proposes a method based on deep unsupervised learning for drug-target interaction prediction called AutoDTI++. The proposed method includes three steps. The first step is to pre-process the interaction matrix. Since the interaction matrix is sparse, we solved the sparsity of the interaction matrix with drug fingerprints. Then, in the second step, the AutoDTI approach is introduced. In the third step, we post-preprocess the output of the AutoDTI model. Conclusions Experimental results have shown that we were able to improve the prediction performance. To this end, the proposed method has been compared to other algorithms using the same reference datasets. The proposed method indicates that the experimental results of running five repetitions of tenfold cross-validation on golden standard datasets (Nuclear Receptors, GPCRs, Ion channels, and Enzymes) achieve good performance with high accuracy.

Digital phenotyping is the term given to the capturing and use of user log data from health and wellbeing technologies used in apps and cloud-based services. This paper explores ethical issues in making use of digital phenotype data in the arena of digital health interventions. Products and services based on digital wellbeing technologies typically include mobile device apps as well as browser-based apps to a lesser extent, and can include telephony-based services, text-based chatbots, and voice-activated chatbots. Many of these digital products and services are simultaneously available across many channels in order to maximize availability for users. Digital wellbeing technologies offer useful methods for real-time data capture of the interactions of users with the products and services. It is possible to design what data are recorded, how and where it may be stored, and, crucially, how it can be analyzed to reveal individual or collective usage patterns. The paper also examines digital phenotyping workflows, before enumerating the ethical concerns pertaining to different types of digital phenotype data, highlighting ethical considerations for collection, storage, and use of the data. A case study of a digital health app is used to illustrate the ethical issues. The case study explores the issues from a perspective of data prospecting and subsequent machine learning. The ethical use of machine learning and artificial intelligence on digital phenotype data and the broader issues in democratizing machine learning and artificial intelligence for digital phenotype data are then explored in detail.

Background Wastage of blood products can be a significant cost to blood banks. However, the cause of wastage is often complex and makes it difficult to determine wastage-associated factors. Machine learning techniques may be useful tools to investigate these complex associations. We investigated whether unsupervised machine learning can identify patterns associated with wastage in our blood bank. Materials and methods Data on red blood cells, platelets and frozen products were obtained from the laboratory information system of the Central Zone Blood Transfusion Services at Nova Scotia Health Authority. A total of 879 532 transactions were analysed by association rule mining, a type of machine learning algorithm. Associations with lift scores greater than 25 and with clinical relevance were flagged for further examination. Results Association rule mining returned a total of 3355 associations related to wastage. Several notable associations were identified. For example, certain wards were associated with wastage due to thawing unused frozen products. Other examples included association between smaller blood banks and evening work shifts with product wastage due to excess time outside the laboratory or returning products with high temperatures. Conclusion This paper demonstrates the effective use of unsupervised machine learning for the purpose of investigating wastage in a large blood bank. The use of association rule mining was able to identify wastage factors, which can help guide quality improvement initiatives. This technique can be automated to provide rapid analysis of complex associations contributing to wastage and could be utilized in modern blood banks.

Objective The characteristics of clinician activities while interacting with electronic health record (EHR) systems can influence the time spent in EHRs and workload. This study aims to characterize EHR activities as tasks and define novel, data-driven metrics. Materials and Methods We leveraged unsupervised learning approaches to learn tasks from sequences of events in EHR audit logs. We developed metrics characterizing the prevalence of unique events and event repetition and applied them to categorize tasks into 4 complexity profiles. Between these profiles, Mann-Whitney U tests were applied to measure the differences in performance time, event type, and clinician prevalence, or the number of unique clinicians who were observed performing these tasks. In addition, we apply process mining frameworks paired with clinical annotations to support the validity of a sample of our identified tasks. We apply our approaches to learn tasks performed by nurses in the Vanderbilt University Medical Center neonatal intensive care unit. Results We examined EHR audit logs generated by 33 neonatal intensive care unit nurses resulting in 57 234 sessions and 81 tasks. Our results indicated significant differences in performance time for each observed task complexity profile. There were no significant differences in clinician prevalence or in the frequency of viewing and modifying event types between tasks of different complexities. We presented a sample of expert-reviewed, annotated task workflows supporting the interpretation of their clinical meaningfulness. Conclusions The use of the audit log provides an opportunity to assist hospitals in further investigating clinician activities to optimize EHR workflows.

Cell morphology encodes essential information on many underlying biological processes. It is commonly used by clinicians and researchers in the study, diagnosis, prognosis, and treatment of human diseases. Quantification of cell morphology has seen tremendous advances in recent years. However, effectively defining morphological shapes and evaluating the extent of morphological heterogeneity within cell populations remain challenging. Here we present a protocol and software for the analysis of cell and nuclear morphology from fluorescence or bright-field images using the VAMPIRE algorithm ( https://github.com/kukionfr/VAMPIRE_open ). This algorithm enables the profiling and classification of cells into shape modes based on equidistant points along cell and nuclear contours. Examining the distributions of cell morphologies across automatically identified shape modes provides an effective visualization scheme that relates cell shapes to cellular subtypes based on endogenous and exogenous cellular conditions. In addition, these shape mode distributions offer a direct and quantitative way to measure the extent of morphological heterogeneity within cell populations. This protocol is highly automated and fast, with the ability to quantify the morphologies from 2D projections of cells seeded both on 2D substrates or embedded within 3D microenvironments, such as hydrogels and tissues. The complete analysis pipeline can be completed within 60 minutes for a dataset of ~20,000 cells/2,400 images.

Methylation profiling has become a mainstay in brain tumor diagnostics since the introduction of the first publicly available classification tool by the German Cancer Research Center in 2017. We demonstrate the capability of this system through an example of a rare case of IDH wildtype glioblastoma diagnosed in a patient previously treated for T-cell acute lymphoblastic leukemia. Our novel in-house diagnostic tool EpiDiP provided hints arguing against a radiation-induced tumor, identified a novel recurrent genetic aberration, and thus informed about a potential therapeutic target.

Composite such as Glass Fibre Reinforced Polymer (GFRP) is increasingly used as insulation in many industrial applications such as the steel pipelines in the oil and gas industry. Due to ageing and cyclic operation, many hidden defects exist under insulation, such as corrosion and delamination. If these defects are not promptly detected and restored, the growth of defects causes a catastrophic loss. Therefore, an effective inspection technique using non-destructive testing (NDT) to detect the underneath defect is required. The ability of microwave signals to penetrate and interact with the inner structure within composites makes them a promising candidate for composite inspection. In the case of GFRP, the random patterns cause permittivity variations that influence the propagation of the microwave signals, which results in a blurred spatial image making the assessment of the material’s state difficult. In this research, a novel microwave NDT technique is presented based on k-means unsupervised machine learning for defect detection in composites. At present, the defect evaluation using an unsupervised machine learning-based microwave NDT technique is not reported elsewhere. The unsupervised machine learning is employed to enhance the imaging efficiency and defect detection in GFRP. The technique is based on scanning the composite material with an open-ended rectangular waveguide operating from 18 to 26.5 GHz with 101 frequency points. The influence of the permittivity variations on the reflected coefficients due to the random patterns of GFRP is mitigated by measuring the mean of a set of the adjacent points at each operating frequency point using a small rectangular window. The measured data is converted to the time domain using a fast inverse Fourier transform (IFFT) to provide significant features and increase the signal resolution to 201-time steps. K-means algorithm is utilized to cluster the given features into the defect and defect-free regions in GFRP. The findings presented in this paper demonstrate the benefits of an unsupervised machine learning to detect a defect down to 1 mm, which is a considerable contribution over any existing defect inspection technique in composites.

Cyber-Physical Systems (CPSs) play a critical role in our modern infrastructure due to their capability to connect computing resources with physical systems. As such, topics such as reliability, performance, and security of CPSs continue to receive increased attention from the research community. CPSs produce massive amounts of data, creating opportunities to use predictive Machine Learning (ML) models for performance monitoring and optimization, preventive maintenance, and threat detection. However, the “black-box” nature of complex ML models is a drawback when used in safety-critical systems such as CPSs. While explainable ML has been an active research area in recent years, much of the work has been focused on supervised learning. As CPSs rapidly produce massive amounts of unlabeled data, relying on supervised learning alone is not sufficient for data-driven decision making in CPSs. Therefore, if we are to maximize the use of ML in CPSs, it is necessary to have explainable unsupervised ML models. In this paper, we outline how unsupervised explainable ML could be used within CPSs. We review the existing work in unsupervised ML, present initial desiderata of explainable unsupervised ML for CPS, and present a Self-Organizing Maps based explainable clustering methodology which generates global and local explanations. We evaluate the fidelity of the generated explanations using feature perturbation techniques. The results show that the proposed method identifies the most important features responsible for the decision-making process of Self-organizing Maps. Further, we demonstrated that explainable Self-Organizing Maps are a strong candidate for explainable unsupervised machine learning by comparing its model capabilities and limitations with current explainable unsupervised methods.

An unsupervised machine learning model of association rule known as market basket analysis is proposed in this study to analyze the influence of various socio-economic factors on the choice of the water source. Data of 51 socio-economic factors collected from 295 individuals living in 65 households in Ambo city in the Oromia region of Ethiopians were used for this purpose. The results revealed (i) 64% of the family preferred multiple water sources (i.e., public tap and river water), (ii) the water was collected females in 92% of the households, and (iii) majority of people preferred bathing and laundering in the river (support = 32% and confidence = 87%). Direct utilization of river water is not a preferable choice for the user since it may lead to severe health issues and cause water pollution from bathing and laundering. Education and monthly income have a significant impact on the choices of water sources. Local management authorities can improve sanitation and public health management using the results obtained in the study. The paper only gives a glimpse of the important factors that should be considered for improving the way of life for the underdeveloped areas of the world using advanced machine learning techniques.

Deep learning analysis of images and text unfolds new horizons in medicine. However, analysis of transcriptomic data, the cause of biological and pathological changes, is hampered by structural complexity distinctive from images and text. Here we conduct unsupervised training on more than 20,000 human normal and tumor transcriptomic data and show that the resulting Deep-Autoencoder, DeepT2Vec, has successfully extracted informative features and embedded transcriptomes into 30-dimensional Transcriptomic Feature Vectors (TFVs). We demonstrate that the TFVs could recapitulate expression patterns and be used to track tissue origins. Trained on these extracted features only, a supervised classifier, DeepC, can effectively distinguish tumors from normal samples with an accuracy of 90% for Pan-Cancer and reach an average 94% for specific cancers. Training on a connected network, the accuracy is further increased to 96% for Pan-Cancer. Together, our study shows that deep learning with autoencoder is suitable for transcriptomic analysis, and DeepT2Vec could be successfully applied to distinguish cancers, normal tissues, and other potential traits with limited samples.

Machine learning algorithms have recently shown their precision and potential in many different use cases and fields of medicine. Most of the algorithms used are supervised and need a large quantity of labeled data to achieve high accuracy. Also, most applications of machine learning in medicine are attempts to mimic or exceed human diagnostic capabilities but little work has been done to show the power of these algorithms to help collect and pre-process a large amount of data. In this study we show how unsupervised learning can extract and sort usable data from large unlabeled datasets with minimal human intervention. Our digital examination tools used in clinical practice store such databases and are largely under-exploited. We applied unsupervised algorithms to corneal topography examinations which remains the gold standard test for diagnosis and follow-up of many corneal diseases and refractive surgery screening. We could extract 7019 usable examinations which were automatically sorted in 3 common diagnoses (Normal, Keratoconus and History of Refractive Surgery) from an unlabeled database with an overall accuracy of 96.5%. Similar methods could be used on any form of digital examination database and greatly speed up the data collection process and yield to the elaboration of stronger supervised models.

Echo planar imaging (EPI) is a fast and non-invasive magnetic resonance imaging technique that supports data acquisition at high spatial and temporal resolutions. However, susceptibility artifacts, which cause the misalignment to the underlying structural image, are unavoidable distortions in EPI. Traditional susceptibility artifact correction (SAC) methods estimate the displacement field by optimizing an objective function that involves one or more pairs of reversed phase-encoding (PE) images. The estimated displacement field is then used to unwarp the distorted images and produce the corrected images. Since this conventional approach is time-consuming, we propose an end-to-end deep learning technique, named S-Net, to correct the susceptibility artifacts the reversed-PE image pair. The proposed S-Net consists of two components: (i) a convolutional neural network to map a reversed-PE image pair to the displacement field; and (ii) a spatial transform unit to unwarp the input images and produce the corrected images. The S-Net is trained using a set of reversed-PE image pairs and an unsupervised loss function, without ground-truth data. For a new image pair of reversed-PE images, the displacement field and corrected images are obtained simultaneously by evaluating the trained S-Net directly. Evaluations on three different datasets demonstrate that S-Net can correct the susceptibility artifacts in the reversed-PE images. Compared with two state-of-the-art SAC methods (TOPUP and TISAC), the proposed S-Net runs significantly faster: 20 times faster than TISAC and 369 times faster than TOPUP, while achieving a similar correction accuracy. Consequently, S-Net accelerates the medical image processing pipelines and makes the real-time correction for MRI scanners feasible. Our proposed technique also opens up a new direction in learning-based SAC.

The heterogeneity of traumatic brain injury (TBI) remains a core challenge for the success of interventional clinical trials. Data-driven approaches for patient stratification may help to identify TBI patient phenotypes during the acute injury period as well as facilitate targeted trial patient enrollment and analysis of treatment efficacy. In this study, we implemented an unsupervised machine learning approach to identify TBI subpopulations at injury baseline using data from 1213 TBI patients who participated in the Citicoline Brain Injury Treatment Trial (COBRIT) Trial. A wrapper framework utilizing generalized low-rank models automatically selected relevant clinical features that were subsequently used to cluster patients using a partitioning around medoids clustering algorithm. Using this approach, we identified three patient phenotypes with unique clinical injury profiles based on a subset of acute injury features. Phenotype-specific differences in long-term functional outcome trajectories were respectively observed at 3 and 6 months after injury. In comparison, when patients were grouped by baseline Glasgow Coma Scale (GCS), no differences in baseline clinical feature profiles or long-term outcomes were observed. To test phenotype reproducibility in an external validation data set, we used a K-nearest neighbors algorithm to classify subjects in the Transforming Research and Clinical Knowledge in Traumatic Brain Injury (TRACK-TBI) Pilot data set into corresponding phenotypes, then measured the Gower's dissimilarities between TRACK-TBI and COBRIT subjects in each phenotype. No significant differences were found between trial subjects within two phenotypes, suggesting that these phenotypes may be generalizable within a broad range of TBI severity. Further, Extended Glasgow Outcome Scale (GOS-E) outcomes in the TRACK-TBI data set similarly demonstrated phenotype-specific differences in long-term outcomes. Our results suggest that unsupervised machine learning is a promising and effective approach for discovery of novel injury subpopulations over the conventional GCS-based method, and may improve patient selection in future TBI clinical trials.

With the development of deep learning in medical image analysis, decoding brain states from functional magnetic resonance imaging (fMRI) signals has made significant progress. Previous studies often utilized deep neural networks to automatically classify brain activity patterns related to diverse cognitive states. However, due to the individual differences between subjects and the variation in acquisition parameters across devices, the inconsistency in data distributions degrades the performance of cross-subject decoding. Besides, most current networks were trained in a supervised way, which is not suitable for the actual scenarios in which massive amounts of data are unlabeled. To address these problems, we proposed the deep cross-subject adaptation decoding (DCAD) framework to decipher the brain states. The proposed volume-based 3D feature extraction architecture can automatically learn the common spatiotemporal features of labeled source data to generate a distinct descriptor. Then, the distance between the source and target distributions is minimized via an unsupervised domain adaptation (UDA) method, which can help to accurately decode the cognitive states across subjects. The performance of the DCAD was evaluated on task-fMRI (tfMRI) dataset from the Human Connectome Project (HCP). Experimental results showed that the proposed method achieved the state-of-the-art decoding performance with mean 81.9% and 84.9% accuracies under two conditions (4 brain states and 9 brain states respectively) of working memory task. Our findings also demonstrated that UDA can mitigate the impact of the data distribution shift, thereby providing a superior choice for increasing the performance of cross-subject decoding without depending on annotations.

We present an Unsupervised Domain Adaptation strategy to compensate for domain shifts on Electron Microscopy volumes. Our method aggregates visual correspondences-motifs that are visually similar across different acquisitions-to infer changes on the parameters of pretrained models, and enable them to operate on new data. In particular, we examine the annotations of an existing acquisition to determine pivot locations that characterize the reference segmentation, and use a patch matching algorithm to find their candidate visual correspondences in a new volume. We aggregate all the candidate correspondences by a voting scheme and we use them to construct a consensus heatmap: a map of how frequently locations on the new volume are matched to relevant locations from the original acquisition. This information allows us to perform model adaptations in two different ways: either by a) optimizing model parameters under a Multiple Instance Learning formulation, so that predictions between reference locations and their sets of correspondences agree, or by b) using high-scoring regions of the heatmap as soft labels to be incorporated in other domain adaptation pipelines, including deep learning ones. We show that these unsupervised techniques allow us to obtain high-quality segmentations on unannotated volumes, qualitatively consistent with results obtained under full supervision, for both mitochondria and synapses, with no need for new annotation effort.

This paper presents an adaptable dictionary-based feature extraction approach for spike sorting offering high accuracy and low computational complexity for implantable applications. It extracts and learns identifiable features from evolving subspaces through matched unsupervised subspace filtering. To provide compatibility with the strict constraints in implantable devices such as the chip area and power budget, the dictionary contains arrays of {–1, 0 and 1} and the algorithm need only process addition and subtraction operations. Three types of such dictionary were considered. To quantify and compare the performance of the resulting three feature extractors with existing systems, a neural signal simulator based on several different libraries was developed. For noise levels σN between 0.05 and 0.3 and groups of 3 to 6 clusters, all three feature extractors provide robust high performance with average classification errors of less than 8% over five iterations, each consisting of 100 generated data segments. To our knowledge, the proposed adaptive feature extractors are the first able to classify reliably 6 clusters for implantable applications. An ASIC implementation of the best performing dictionary-based feature extractor was synthesized in a 65-nm CMOS process. It occupies an area of 0.09 mm2 and dissipates up to about 10.48 μW from a 1 V supply voltage, when operating with 8-bit resolution at 30 kHz operating frequency.

Chromatin interaction studies can reveal how the genome is organized into spatially confined sub-compartments in the nucleus. However, accurately identifying sub-compartments from chromatin interaction data remains a challenge in computational biology. Here, we present Sub-Compartment Identifier (SCI), an algorithm that uses graph embedding followed by unsupervised learning to predict sub-compartments using Hi-C chromatin interaction data. We find that the network topological centrality and clustering performance of SCI sub-compartment predictions are superior to those of hidden Markov model (HMM) sub-compartment predictions. Moreover, using orthogonal Chromatin Interaction Analysis by in-situ Paired-End Tag Sequencing (ChIA-PET) data, we confirmed that SCI sub-compartment prediction outperforms HMM. We show that SCI-predicted sub-compartments have distinct epigenetic marks, transcriptional activities, and transcription factor enrichment. Moreover, we present a deep neural network to predict sub-compartments using epigenome, replication timing, and sequence data. Our neural network predicts more accurate sub-compartment predictions when SCI-determined sub-compartments are used as labels for training., Accurate identification of sub-compartments from chromatin interaction data remains a challenge. Here, the authors introduce an algorithm combining graph embedding and unsupervised learning to predict sub-compartments using Hi-C data.

Wind power experienced a substantial growth over the past decade especially because it has been seen as one of the best ways towards meeting climate change and emissions targets by many countries. Since wind power is not fully dispatchable, the accuracy of wind forecasts is a key element for the electric system operators, as it strongly affects the decision-making processes. The planning horizon can be short term (1 -3 months) and long-term (6–12 months) depending on the process.The objective of this paper is to conduct a performance comparison of five deep learning models each combined with three types of data pre-processing and used for short term and long-term multi-variate predictions. The input data are time series of the wind power capacity factor and the temperature. In addition, this paper sets out to demonstrate and review the state-of-the-art deep learning models for prediction with a secondary objective to present the reader a reference point to better understand which model to choose and what factors are significant. The first contribution of this paper is to apply, assess and compare a selection of the novel and cutting-edge deep learning models for multi-variate prediction. Multi-variate predication is achieved through a proposed multiple input and multiple output (MIMO) architecture. Compared to traditional prediction models, machine learning techniques have the advantage of generalization. Among various techniques deep learning is particularly getting more attention due to the applicability to various dataset such as numerical and character. This investigation focuses on five models — Deep Feed Forward (DFF), Deep Convolutional Network (DCN), Recurrent Neural Network (RNN), Attention mechanism (Attention) and Long Short-Term Memory Networks (LSTM). The second contribution is to propose a novel approach to transform the time series dataset to signal for input and reconstruct the model predictions through inverse transformation, by means of the so-called discrete wavelet transformation and fast Fourier transformation. The different models are assessed also by comparing their performance with and without the input dataset manipulation through wavelet and FFT transformation. Beyond that, the model performances are outlined in detail, to give the reader an overview of the models to choose from for short-term or long-term prediction. The results demonstrate that the Attention and DCN perform best with Wavelet or FFT signal, whereas some other models perform better with no data preprocessing. Keywords: Unsupervised machine learning, Multi-variate prediction, Short and long term prediction, Wind power forecasting, Time series to frequency transformation, Performance evaluation and comparison

BACKGROUND CONTEXT Traumatic spinal cord injury can have a dramatic effect on a patient's life. The degree of neurologic recovery greatly influences a patient's treatment and expected quality of life. This has resulted in the development of machine learning algorithms (MLA) that use acute demographic and neurologic information to prognosticate recovery. The van Middendorp et al. (2011) (vM) logistic regression (LR) model has been established as a reference model for the prediction of walking recovery following spinal cord injury as it has been validated within many different countries. However, an examination of the way in which these prediction models are evaluated is warranted. The area under the receiver operators curve (AUROC) has been consistently used when evaluating model performance, but it has been shown that AUROC overemphasizes the most common event resulting in an inaccurate assessment when the data are imbalanced. Furthermore, there is evidence that the use of more advanced MLA, such as an unsupervised k-means model, may show superior performance compared to LR as they can handle a larger number of features. PURPOSE The first objective of the study was to assess the performance of both an unsupervised MLA and LR model with complete admission neurologic information against the vM and Hicks models. Second, a comparison between the accuracy of the AUROC and the F1-score will be made to determine which method is superior for the assessment of diagnostic performance of prediction models on large-scale datasets. STUDY DESIGN Retrospective review of a prospective cohort study. PATIENT SAMPLE The Rick Hansen Spinal Cord Injury Registry (RHSCIR) was used in this study. All patients enrolled between 2004 and 2017 with complete neurologic examination and Functional Independence Measure outcome data at ≥1 year follow-up or who could walk at discharge were included. The prognostic variables included age (dichotomized at ≥65 years old); American Spinal Injury Association Impairment Scale (AIS) grade; and individual motor, light touch, and pinprick score from L2 to S1. OUTCOME MEASURES The Functional Independence Measure locomotor score was used to assess independent walking ability at discharge or 1-year follow-up. METHODS An unsupervised MLA with k=2 was chosen in order to identify a “walk” cluster and a “not walk” cluster. Model performance was assessed through the development of a receiver operating characteristic curve with associated AUROC and a precision-recall curve with associated F1-score. The study and the RHSCIR are supported by funding from Health Canada, Western Economic Diversification Canada, and the Governments of Alberta, British Columbia, Manitoba, and Ontario. These funders had no role in the study or study reporting and the authors have no conflicts of interest to report. RESULTS No clinically relevant differences were found between with the use of an unsupervised MLA with a greater amount of initial neurologic information compared to the established standards for any AIS classification. Although demonstrated for all separate AIS classifications, most notably, the AUROC for the vM (0.78) and Hicks models (0.76) were found to be superior to that of the new LR model (0.72); however, the vM and Hicks models had more than double the amount of false negative classifications compared to the LR. The F1-scores between these three models were also found to be different but with the vM and Hicks models being lower than the LR (0.85, 0.81, and 0.89, respectively). CONCLUSIONS No clinically relevant differences were found between the use of an unsupervised MLA with complete admission neurologic information compared to the previously validated standards; however, when comparing the performance of the AUROC and F1-score, the AUROC showed inaccurate prognostic performance when there was an imbalance toward a greater amount of false negatives. Importantly, the F1-score did not succumb to this imbalance. As AUROC has been used as the standard when evaluating performance of prediction models, consideration as to whether this is the most appropriate method is warranted. Future work should focus on comparing AUROC and F1-scores with other previously validated models.

This study applies adaptive mixture independent component analysis (AMICA) to learn a set of ICA models, each optimized by fitting a distributional model for each identified component process while maximizing component process independence within some subsets of time points of a multi-channel EEG dataset. Here, we applied 20-model AMICA decomposition to long-duration (1-2h), high-density (128-channel) EEG data recorded while participants used guided imagination to imagine situations stimulating the experience of 15 specified emotions. These decompositions tended to return models identifying spatiotemporal EEG patterns or states within single emotion imagination periods. Model probability transitions reflected time-courses of EEG dynamics during emotion imagination, which varied across emotions. Transitions between models accounting for imagined "grief" and "happiness" were more abrupt and better aligned with participant reports, while transitions for imagined "contentment" extended into adjoining "relaxation" periods. The spatial distributions of brain-localizable independent component processes (ICs) were more similar within participants (across emotions) than emotions (across participants). Across participants, brain regions with differences in IC spatial distributions (i.e., dipole density) between emotion imagination versus relaxation were identified in or near the left rostrolateral prefrontal, posterior cingulate cortex, right insula, bilateral sensorimotor, premotor, and associative visual cortex. No difference in dipole density was found between positive versus negative emotions. AMICA models of changes in high-density EEG dynamics may allow data-driven insights into brain dynamics during emotional experience, possibly enabling the improved performance of EEG-based emotion decoding and advancing our understanding of emotion.

Among the dairy sector's current concerns, the as-sessment of global animal health status is a complex challenge. Its multidimensionality means that global monitoring tools are rarely considered. Instead, specific disease detection is often studied separately and, due to financial and ethical issues, uses small-scale data sets focusing on few biomarkers. Several studies have already been conducted using milk Fourier transform mid-infrared (FT-MIR) spectroscopy to detect mastitis and lameness or to quantify health-related biomarkers in milk or blood. Those studies are relevant but they focus mainly on one biomarker or disease. To solve this issue and the small-scale data set, in this study, we proposed a holistic approach using big data obtained from milk recording, including milk yield, somatic cell count, and 27 FT-MIR-based predictors related to milk composition and animal health status. Using 740,454 records collected from 114,536 first-parity Holstein cows in southern Belgium, we performed repeated unsupervised learning algorithms based on Ward's agglomerative hierarchical clustering method to find potential interesting patterns. A divide-and-conquer approach was used to overcome the limitation of computational resources in clustering a relatively large data set. Five groups of records were identified. Differences observed in the fourth group suggested a relationship to metabolic disorders. The fifth group seemed to be related to mastitis. In a second step, we performed a partial least squares discriminant analysis (PLS-DA) to predict the probability of belonging to those specific groups for the entire data set. The obtained global ac-curacy was 0.77 and the balanced accuracy (i.e., the mean between sensitivity and specificity) of discrimi-nating the fourth and fifth groups was 0.88 and 0.96, respectively. Then, a validation of the interpretation of those groups was performed using 204 milk and blood reference records. The predicted probability associated with the metabolic disorders issue had significant cor-relations of 0.54 with blood beta-hydroxybutyrate, 0.44 with blood nonesterified fatty acids, -0.32 with blood glucose, -0.23 with milk glucose-6-phosphate, and 0.38 with milk isocitrate. In contrast, the predicted prob-ability of belonging to the mastitis group had correla-tions of 0.69 with milk lactate dehydrogenase, 0.46 with milk N-acetyl-beta-D-glucosaminidase, -0.18 with milk free glucose, and 0.16 with milk glucose-6-phosphate. Consequently, these results suggest that the obtained quantitative traits indirectly reflect some of the main health disorders in dairy farming and could be used to monitor dairy cows on a large scale. By using unsuper-vised learning on large-scale milk recording data and then validating the pattern using reference laboratory measures, we propose a new approach to quickly assess dairy cow health status.

The availability of computers has brought novel prospects in drug design. Neural networks (NN) were an early tool that cheminformatics tested for converting data into drugs. However, the initial interest faded for almost two decades. The recent success of Deep Learning (DL) has inspired a renaissance of neural networks for their potential application in deep chemistry. DL targets direct data analysis without any human intervention. Although back-propagation NN is the main algorithm in the DL that is currently being used, unsupervised learning can be even more efficient. We review self-organizing maps (SOM) in mapping molecular representations from the 1990s to the current deep chemistry. We discovered the enormous efficiency of SOM not only for features that could be expected by humans, but also for those that are not trivial to human chemists. We reviewed the DL projects in the current literature, especially unsupervised architectures. DL appears to be efficient in pattern recognition (Deep Face) or chess (Deep Blue). However, an efficient deep chemistry is still a matter for the future. This is because the availability of measured property data in chemistry is still limited.

The power grid is a build-up of a mesh of thousands of sensors, embedded devices, and terminal units that communicate over different media. The heterogeneity of modern and legacy equipment calls for attention towards diverse network security measures. The critical infrastructure employs different security measures to detect and prevent adversaries, e.g., through signature-based tools. These approaches lack the potential to identify unknown attacks. Machine learning has the prospective to address novel attack vectors. This paper systematically evaluates the efficacy of learning algorithms from different families for intrusion detection in IEC 60870-5-104 protocol. One-class SVM and k-Nearest Neighbour unsupervised learning models show small potential when being tested on the IEC 104 unseen dataset with Area Under the Curve score 0.64 and 0.59, in the same order; and Matthews Correlation Coefficient value 0.3 and 0.2, respectively. The experimental results suggest little feasibility of the evaluated unsupervised learning approaches for anomaly detection in IEC 104 communication and recommend coupling it with other anomaly detection techniques. © 2021 IEEE.

Aims: This study presents a survival stratification model based on multi-omics integration using bidirectional deep neural networks (BiDNNs) in gastric cancer. Methods: Based on the survival-related representation features yielded by BiDNNs through integrating transcriptomics and epigenomics data, K-means clustering analysis was performed to cluster tumor samples into different survival subgroups. The BiDNNs-based model was validated using tenfold cross-validation and in two independent confirmation cohorts. Results: Using the BiDNNs-based survival stratification model, patients were grouped into two survival subgroups with log-rank p-value = 9.05E-05. The subgroups classification was robustly validated in tenfold cross-validation (C-index = 0.65 ± 0.02) and in two confirmation cohorts (E-GEOD-26253, C-index = 0.609; E-GEOD-62254, C-index = 0.706). Conclusion: We propose and validate a robust and stable BiDNN-based survival stratification model in gastric cancer.