Search

Abstract Background Aneuploidy, an abnormal number of chromosomes within a cell, is a hallmark of cancer. Patterns of aneuploidy differ across cancers, yet are similar in cancers affecting closely related tissues. The selection pressures underlying aneuploidy patterns are not fully understood, hindering our understanding of cancer development and progression. Results Here, we apply interpretable machine learning methods to study tissue-selective aneuploidy patterns. We define 20 types of features corresponding to genomic attributes of chromosome-arms, normal tissues, primary tumors, and cancer cell lines (CCLs), and use them to model gains and losses of chromosome arms in 24 cancer types. To reveal the factors that shape the tissue-specific cancer aneuploidy landscapes, we interpret the machine learning models by estimating the relative contribution of each feature to the models. While confirming known drivers of positive selection, our quantitative analysis highlights the importance of negative selection for shaping aneuploidy landscapes. This is exemplified by tumor suppressor gene density being a better predictor of gain patterns than oncogene density, and vice versa for loss patterns. We also identify the importance of tissue-selective features and demonstrate them experimentally, revealing KLF5 as an important driver for chr13q gain in colon cancer. Further supporting an important role for negative selection in shaping the aneuploidy landscapes, we find compensation by paralogs to be among the top predictors of chromosome arm loss prevalence and demonstrate this relationship for one paralog interaction. Similar factors shape aneuploidy patterns in human CCLs, demonstrating their relevance for aneuploidy research. Conclusions Our quantitative, interpretable machine learning models improve the understanding of the genomic properties that shape cancer aneuploidy landscapes.

MicroRNAs (miRNAs) play a crucial role in mRNA regulation. Identifying functionally important mRNA targets of a specific miRNA is essential for uncovering its biological function and assisting miRNA–based drug development. Datasets of high-throughput direct bona fide miRNA–target interactions (MTIs) exist only for a few model organisms, prompting the need for computational prediction. However, the scarcity of data poses a challenge in training accurate machine learning models for MTI prediction. In this study, we explored the potential of transfer learning technique (with ANN and XGB) to address the limited data challenge by leveraging the similarities in interaction rules between species. Furthermore, we introduced a novel approach called TransferSHAP for estimating the feature importance of transfer learning in tabular dataset tasks. We demonstrated that transfer learning improves MTI prediction accuracy for species with limited datasets and identified the specific interaction features the models employed to transfer information across different species.

Abstract Generating novel valid molecules is often a difficult task, because the vast chemical space relies on the intuition of experienced chemists. In recent years, deep learning models have helped accelerate this process. These advanced models can also help identify suitable molecules for disease treatment. In this paper, we propose Taiga, a transformer-based architecture for the generation of molecules with desired properties. Using a two-stage approach, we first treat the problem as a language modeling task of predicting the next token, using SMILES strings. Then, we use reinforcement learning to optimize molecular properties such as QED. This approach allows our model to learn the underlying rules of chemistry and more easily optimize for molecules with desired properties. Our evaluation of Taiga, which was performed with multiple datasets and tasks, shows that Taiga is comparable to, or even outperforms, state-of-the-art baselines for molecule optimization, with improvements in the QED ranging from 2 to over 20 percent. The improvement was demonstrated both on datasets containing lead molecules and random molecules. We also show that with its two stages, Taiga is capable of generating molecules with higher biological property scores than the same model without reinforcement learning.

Introduction Atrial fibrillation (AF) is a major public health issue and there is rationale for the early diagnosis of AF before the first complication occurs. Previous AF screening research is limited by low yields of new cases and strokes prevented in the screened populations. For AF screening to be clinically and cost-effective, the efficiency of identification of newly diagnosed AF needs to be improved and the intervention offered may have to extend beyond oral anticoagulation for stroke prophylaxis. Previous prediction models for incident AF have been limited by their data sources and methodologies.Methods and analysis We will investigate the application of random forest and multivariable logistic regression to predict incident AF within a 6-month prediction horizon, that is, a time-window consistent with conducting investigation for AF. The Clinical Practice Research Datalink (CPRD)-GOLD dataset will be used for derivation, and the Clalit Health Services (CHS) dataset will be used for international external geographical validation. Analyses will include metrics of prediction performance and clinical utility. We will create Kaplan-Meier plots for individuals identified as higher and lower predicted risk of AF and derive the cumulative incidence rate for non-AF cardio-renal-metabolic diseases and death over the longer term to establish how predicted AF risk is associated with a range of new non-AF disease states.Ethics and dissemination Permission for CPRD-GOLD was obtained from CPRD (ref no: 19_076). The CPRD ethical approval committee approved the study. CHS Helsinki committee approval 21-0169 and data usage committee approval 901. The results will be submitted as a research paper for publication to a peer-reviewed journal and presented at peer-reviewed conferences.Trial registration number A systematic review to guide the overall project was registered on PROSPERO (registration number CRD42021245093). The study was registered on ClinicalTrials.gov (NCT05837364).

Abstract How do aberrations in widely expressed genes lead to tissue‐selective hereditary diseases? Previous attempts to answer this question were limited to testing a few candidate mechanisms. To answer this question at a larger scale, we developed “Tissue Risk Assessment of Causality by Expression” (TRACE), a machine learning approach to predict genes that underlie tissue‐selective diseases and selectivity‐related features. TRACE utilized 4,744 biologically interpretable tissue‐specific gene features that were inferred from heterogeneous omics datasets. Application of TRACE to 1,031 disease genes uncovered known and novel selectivity‐related features, the most common of which was previously overlooked. Next, we created a catalog of tissue‐associated risks for 18,927 protein‐coding genes (https://netbio.bgu.ac.il/trace/). As proof‐of‐concept, we prioritized candidate disease genes identified in 48 rare‐disease patients. TRACE ranked the verified disease gene among the patient's candidate genes significantly better than gene prioritization methods that rank by gene constraint or tissue expression. Thus, tissue selectivity combined with machine learning enhances genetic and clinical understanding of hereditary diseases.

Abstract Background Drug–drug interactions (DDIs) are preventable causes of medical injuries and often result in doctor and emergency room visits. Previous research demonstrates the effectiveness of using matrix completion approaches based on known drug interactions to predict unknown Drug–drug interactions. However, in the case of a new drug, where there is limited or no knowledge regarding the drug’s existing interactions, such an approach is unsuitable, and other drug’s preferences can be used to accurately predict new Drug–drug interactions. Methods We propose adjacency biomedical text embedding (ABTE) to address this limitation by using a hybrid approach which combines known drugs’ interactions and the drug’s biomedical text embeddings to predict the DDIs of both new and well known drugs. Results Our evaluation demonstrates the superiority of this approach compared to recently published DDI prediction models and matrix factorization-based approaches. Furthermore, we compared the use of different text embedding methods in ABTE, and found that the concept embedding approach, which involves biomedical information in the embedding process, provides the highest performance for this task. Additionally, we demonstrate the effectiveness of leveraging biomedical text embedding for additional drugs’ biomedical prediction task by presenting text embedding’s contribution to a multi-modal pregnancy drug safety classification. Conclusion Text and concept embeddings created by analyzing a domain-specific large-scale biomedical corpora can be used for predicting drug-related properties such as Drug–drug interactions and drug safety prediction. Prediction models based on the embeddings resulted in comparable results to hand-crafted features, however text embeddings do not require manual categorization or data collection and rely solely on the published literature.

Measurement(s) drug combination effect modeling • drug combination effect modeling Technology Type(s) Text mining • Clinical Trials Informatics System Factor Type(s) Medicine Sample Characteristic - Organism Homo sapiens

Drug-drug interactions (DDIs) are a critical component of drug safety surveillance. Laboratory studies aimed at detecting DDIs are typically difficult, expensive, and time-consuming; therefore, developing in-silico methods is critical. Machine learning-based approaches for DDI prediction have been developed; however, in many cases, their ability to achieve high accuracy relies on data only available towards the end of the molecule lifecycle. Here, we propose a simple yet effective similarity-based method for preclinical DDI prediction where only the chemical structure is available. We test the model on new, unseen drugs. To focus on the preclinical problem setting, we conducted a retrospective analysis and tested the models on drugs that were added to a later version of the DrugBank database. We extend an existing method, adjacency matrix factorization with propagation (AMFP), to support unseen molecules by applying a new lookup mechanism to the drugs' chemical structure, lookup adjacency matrix factorization with propagation (LAMFP). We show that using an ensemble of different similarity measures improves the results. We also demonstrate that Chemprop, a message-passing neural network, can be used for DDI prediction. In computational experiments, LAMFP results in high accuracy, with an area under the receiver operating characteristic curve of 0.82 for interactions involving a new drug and an existing drug and for interactions involving only existing drugs. Moreover, LAMFP outperforms state-of-the-art, complex graph neural network DDI prediction methods.

Machine learning-based Network Intrusion Detection Systems (NIDS) are designed to protect networks by identifying anomalous behaviors or improper uses. In recent years, advanced attacks, such as those mimicking legitimate traffic, have been developed to avoid alerting such systems. Previous works mainly focused on improving the anomaly detector itself, whereas in this paper, we introduce a novel method, Test-Time Augmentation for Network Anomaly Detection (TTANAD), which utilizes test-time augmentation to enhance anomaly detection from the data side. TTANAD leverages the temporal characteristics of traffic data and produces temporal test-time augmentations on the monitored traffic data. This method aims to create additional points of view when examining network traffic during inference, making it suitable for a variety of anomaly detector algorithms. Our experimental results demonstrate that TTANAD outperforms the baseline in all benchmark datasets and with all examined anomaly detection algorithms, according to the Area Under the Receiver Operating Characteristic (AUC) metric.

Highlights Novel artificial nose based upon electrode-deposited carbon dots (C-dots). Significant selectivity and sensitivity determined by “polarity matching” between the C-dots and gas molecules. The C-dot artificial nose facilitates, for the first time, real-time, continuous monitoring of bacterial proliferation and discrimination among bacterial species, both between Gram-positive and Gram-negative bacteria and between specific strains. Machine learning algorithm furnishes excellent predictability both in the case of individual gases and for complex gas mixtures. Abstract Continuous, real-time monitoring and identification of bacteria through detection of microbially emitted volatile molecules are highly sought albeit elusive goals. We introduce an artificial nose for sensing and distinguishing vapor molecules, based upon recording the capacitance of interdigitated electrodes (IDEs) coated with carbon dots (C-dots) exhibiting different polarities. Exposure of the C-dot-IDEs to volatile molecules induced rapid capacitance changes that were intimately dependent upon the polarities of both gas molecules and the electrode-deposited C-dots. We deciphered the mechanism of capacitance transformations, specifically substitution of electrode-adsorbed water by gas molecules, with concomitant changes in capacitance related to both the polarity and dielectric constants of the vapor molecules tested. The C-dot-IDE gas sensor exhibited excellent selectivity, aided by application of machine learning algorithms. The capacitive C-dot-IDE sensor was employed to continuously monitor microbial proliferation, discriminating among bacteria through detection of distinctive “volatile compound fingerprint” for each bacterial species. The C-dot-IDE platform is robust, reusable, readily assembled from inexpensive building blocks and constitutes a versatile and powerful vehicle for gas sensing in general, bacterial monitoring in particular.

This work presents a novel method for applying test-time augmentation (TTA) to tabular data. We used TTA along with an ensemble of 42 models to achieve higher performance on the MIT Global Open Source Severity of Illness Score dataset consisting of 131,051 ICU visits and outcomes. This method achieved an AUC of 0.915 on the private test set (19,669 admissions) and won first place at Stanford University’s WiDS Datathon 2020 challenge on Kaggle, while the Acute Physiology and Chronic Health Evaluation (APACHE) IV model (commonly used for ICU survival prediction in the literature) achieved an AUC of 0.868. In addition to increasing the AUC score, our method also reduces “unfair” bias.

Transfer Learning aims to transfer knowledge from a source task to a target task. We focus on a situation when there is a large number of available source models, and we are interested in choosing a single source model that can maximize the predictive performance in the target domain. Existing methods compute some form of “similarity” between the source task data and the target task data. They then select the most similar source task and use the model trained on it for transfer learning. Previous methods do not account for the fact that it is the model parameters that are transferred rather than the data. Therefore, the “similarity” of the source data does not directly influence transfer learning performance. In addition, we would like the possibility of confidently selecting a source model even when the data it was trained on is not available, for example, due to privacy or copyright constraints. We propose to use the truncated source models as encoders for the target data. We then select a source model based on how well it clusters the target data in the latent encoding space, which we calculate using the Mean Silhouette Coefficient. We prove that if the encodings achieve a Mean Silhouette Coefficient of 1, optimal classification can be achieved using just the final layer of the target network. We evaluate our method using the University of California, Riverside (UCR) time series archive and show that the proposed method achieves comparable results to previous work, without using the source data.

Patients with chronic lymphocytic leukemia (CLL) have a suboptimal humoral response to vaccination. Recently, BNT162b2, an mRNA COVID-19 vaccine with a high efficacy of 95% in immunocompetent individuals, was introduced. We investigated the safety and efficacy of the BNT162b2 mRNA COVID-19 vaccine in patients with CLL from nine medical centers in Israel, Overall 400 patients were included, of whom 373 were found to be eligible for the analysis of antibody response. The vaccine appeared to be safe and only grade 1-2 adverse events were seen in 50% of the patients. Following the second dose, an antibody response was detected in 43% of the cohort. Among these CLL patients, 61% of the treatment-na ve patients responded to the vaccine, while responses developed in only 18% of those with ongoing disease, 37% of those previously treated with a BTK inhibitor and 5% of those recently given an anti-CD20 antibody. Among patients treated with BCL2 as monotherapy or in combination with anti-CD20, 62% and 14%, respectively, developed an immune response. There was a high concordance between neutralizing antibodies and positive serological response to spike protein. Based on our findings we developed a simple seven-factor score including timing of any treatment with anti-CD20, age, treatment status, and IgG, IgA, IgM and hemoglobin levels. The sum of all the above parameters can serve as a possible estimate to predict whether a given CLL patient will develop sufficient antibodies. In conclusion, the BNT162b2 mRNA COVID-19 vaccine was found to be safe in patients with CLL, but its efficacy is limited, particularly in treated patients.

Researchers from academia and the corporate-sector rely on scholarly digital libraries to access articles. Attackers take advantage of innocent users who consider the articles' files safe and thus open PDF-files with little concern. In addition, researchers consider scholarly libraries a reliable, trusted, and untainted corpus of papers. For these reasons, scholarly digital libraries are an attractive-target and inadvertently support the proliferation of cyber-attacks launched via malicious PDF-files. In this study, we present related vulnerabilities and malware distribution approaches that exploit the vulnerabilities of scholarly digital libraries. We evaluated over two-million scholarly papers in the CiteSeerX library and found the library to be contaminated with a surprisingly large number (0.3-2%) of malicious PDF documents (over 55% were crawled from the IPs of US-universities). We developed a two layered detection framework aimed at enhancing the detection of malicious PDF documents, Sec-Lib, which offers a security solution for large digital libraries. Sec-Lib includes a deterministic layer for detecting known malware, and a machine learning based layer for detecting unknown malware. Our evaluation showed that scholarly digital libraries can detect 96.9% of malware with Sec-Lib, while minimizing the number of PDF-files requiring labeling, and thus reducing the manual inspection efforts of security-experts by 98%.

Driving under the influence of alcohol is a widespread phenomenon in the US where it is considered a major cause of fatal accidents. In this research, we present Virtual Breathalyzer, a novel approach for detecting intoxication from the measurements obtained by the sensors of smartphones and wrist-worn devices. We formalize the problem of intoxication detection as the supervised machine learning task of binary classification (drunk or sober). In order to evaluate our approach, we conducted a field experiment and collected 60 free gait samples from 30 patrons of three bars using a Microsoft Band and Samsung Galaxy S4. We validated our results against an admissible breathalyzer used by the police. A system based on this concept successfully detected intoxication and achieved the following results: 0.97 AUC and 0.04 FPR, given a fixed TPR of 1.0. Our approach can be used to analyze the free gait of drinkers when they walk from the car to the bar and vice versa, using wearable devices which are ubiquitous and more widespread than admissible breathalyzers. This approach can be utilized to alert people, or even a connected car, and prevent people from driving under the influence of alcohol.

Drug-drug interactions are preventable causes of medical injuries and often result in doctor and emergency room visits. Computational techniques can be used to predict potential drug-drug interactions. We approach the drug-drug interaction prediction problem as a link prediction problem and present two novel methods for drug-drug interaction prediction based on artificial neural networks and factor propagation over graph nodes: adjacency matrix factorization (AMF) and adjacency matrix factorization with propagation (AMFP). We conduct a retrospective analysis by training our models on a previous release of the DrugBank database with 1,141 drugs and 45,296 drug-drug interactions and evaluate the results on a later version of DrugBank with 1,440 drugs and 248,146 drug-drug interactions. Additionally, we perform a holdout analysis using DrugBank. We report an area under the receiver operating characteristic curve score of 0.807 and 0.990 for the retrospective and holdout analyses respectively. Finally, we create an ensemble-based classifier using AMF, AMFP, and existing link prediction methods and obtain an area under the receiver operating characteristic curve of 0.814 and 0.991 for the retrospective and the holdout analyses. We demonstrate that AMF and AMFP provide state of the art results compared to existing methods and that the ensemble-based classifier improves the performance by combining various predictors. Additionally, we compare our methods with multi-source data-based predictors using cross-validation. In the multi-source data comparison, our methods outperform various ensembles created using 29 different predictors based on several data sources. These results suggest that AMF, AMFP, and the proposed ensemble-based classifier can provide important information during drug development and regarding drug prescription given only partial or noisy data. Additionally, the results indicate that the interaction network (known DDIs) is the most useful data source for identifying potential DDIs and that our methods take advantage of it better than the other methods investigated. The methods we present can also be used to solve other link prediction problems. Drug embeddings (compressed representations) created when training our models using the interaction network have been made public.

Bearing spall detection and predicting its size are great challenges. Model-based simulation is a well-known traditional approach to physically model the influence of the spall on the bearing. Building a physical model is challenging due to the bearing complexity and the expert knowledge required to build such a model. Obviously, building a partial physical model for some of the spall sizes is easier. In this paper, we propose a machine-learning algorithm, called Probability-Based Forest, that uses a partial physical model. First, the behavior of some of the spall sizes is physically modeled and a simulator based on this model generates scenarios for these spall sizes in different conditions. Then, the machine-learning algorithm trains these scenarios to generate a prediction model of spall sizes even for those that have not been modeled by the physical model. Feature extraction is a key factor in the success of this approach. We extract features using two traditional approaches: statistical and physical, and an additional new approach: Time Series FeatuRe Extraction based on Scalable Hypothesis tests (TSFRESH). Experimental evaluation with well-known physical model shows that our approach achieves high accuracy, even in cases that have not been modeled by the physical model. Also, we show that the TSFRESH feature-extraction approach achieves the highest accuracy.

[This corrects the article DOI: 10.1371/journal.pone.0205499.].

Single-cell RNA sequencing (scRNA-seq) is an emerging technology for profiling the gene expression of thousands of cells at the single cell resolution. Currently, the labeling of cells in an scRNA-seq dataset is performed by manually characterizing clusters of cells or by fluorescence-activated cell sorting (FACS). Both methods have inherent drawbacks: The first depends on the clustering algorithm used and the knowledge and arbitrary decisions of the annotator, and the second involves an experimental step in addition to the sequencing and cannot be incorporated into the higher throughput scRNA-seq methods. We therefore suggest a different approach for cell labeling, namely, classifying cells from scRNA-seq datasets by using a model transferred from different (previously labeled) datasets. This approach can complement existing methods, and-in some cases-even replace them. Such a transfer-learning framework requires selecting informative features and training a classifier. The specific implementation for the framework that we propose, designated ''CaSTLe-classification of single cells by transfer learning,'' is based on a robust feature engineering workflow and an XGBoost classification model built on these features. Evaluation of CaSTLe against two benchmark feature-selection and classification methods showed that it outperformed the benchmark methods in most cases and yielded satisfactory classification accuracy in a consistent manner. CaSTLe has the additional advantage of being parallelizable and well suited to large datasets. We showed that it was possible to classify cell types using transfer learning, even when the databases contained a very small number of genes, and our study thus indicates the potential applicability of this approach for analysis of scRNA-seq datasets.