On the effectiveness of compact biomedical transformers.

MOTIVATION: Language models pre-trained on biomedical corpora, such as BioBERT, have recently shown promising results on downstream biomedical tasks. Many existing pre-trained models, on the other hand, are resource-intensive and computationally heavy owing to factors such as embedding size, hidden dimension and number of layers. The natural language processing community has developed numerous strategies to compress these models utilizing techniques such as pruning, quantization and knowledge distillation, resulting in models that are considerably faster, smaller and subsequently easier to use in practice. By the same token, in this article, we introduce six lightweight models, namely, BioDistilBERT, BioTinyBERT, BioMobileBERT, DistilBioBERT, TinyBioBERT and CompactBioBERT which are obtained either by knowledge distillation from a biomedical teacher or continual learning on the Pubmed dataset. We evaluate all of our models on three biomedical tasks and compare them with BioBERT-v1.1 to create the best efficient lightweight models that perform on par with their larger counterparts. RESULTS: We trained six different models in total, with the largest model having 65 million in parameters and the smallest having 15 million; a far lower range of parameters compared with BioBERT's 110M. Based on our experiments on three different biomedical tasks, we found that models distilled from a biomedical teacher and models that have been additionally pre-trained on the PubMed dataset can retain up to 98.8% and 98.6% of the performance of the BioBERT-v1.1, respectively. Overall, our best model below 30 M parameters is BioMobileBERT, while our best models over 30 M parameters are DistilBioBERT and CompactBioBERT, which can keep up to 98.2% and 98.8% of the performance of the BioBERT-v1.1, respectively. AVAILABILITY AND IMPLEMENTATION: Codes are available at: https://github.com/nlpie-research/Compact-Biomedical-Transformers. Trained models can be accessed at: https://huggingface.co/nlpie.

Machine learning to predict the source of campylobacteriosis using whole genome data.

Campylobacteriosis is among the world's most common foodborne illnesses, caused predominantly by the bacterium Campylobacter jejuni. Effective interventions require determination of the infection source which is challenging as transmission occurs via multiple sources such as contaminated meat, poultry, and drinking water. Strain variation has allowed source tracking based upon allelic variation in multi-locus sequence typing (MLST) genes allowing isolates from infected individuals to be attributed to specific animal or environmental reservoirs. However, the accuracy of probabilistic attribution models has been limited by the ability to differentiate isolates based upon just 7 MLST genes. Here, we broaden the input data spectrum to include core genome MLST (cgMLST) and whole genome sequences (WGS), and implement multiple machine learning algorithms, allowing more accurate source attribution. We increase attribution accuracy from 64% using the standard iSource population genetic approach to 71% for MLST, 85% for cgMLST and 78% for kmerized WGS data using the classifier we named aiSource. To gain insight beyond the source model prediction, we use Bayesian inference to analyse the relative affinity of C. jejuni strains to infect humans and identified potential differences, in source-human transmission ability among clonally related isolates in the most common disease causing lineage (ST-21 clonal complex). Providing generalizable computationally efficient methods, based upon machine learning and population genetics, we provide a scalable approach to global disease surveillance that can continuously incorporate novel samples for source attribution and identify fine-scale variation in transmission potential.

Addressing label noise for electronic health records: insights from computer vision for tabular data.

The analysis of extensive electronic health records (EHR) datasets often calls for automated solutions, with machine learning (ML) techniques, including deep learning (DL), taking a lead role. One common task involves categorizing EHR data into predefined groups. However, the vulnerability of EHRs to noise and errors stemming from data collection processes, as well as potential human labeling errors, poses a significant risk. This risk is particularly prominent during the training of DL models, where the possibility of overfitting to noisy labels can have serious repercussions in healthcare. Despite the well-documented existence of label noise in EHR data, few studies have tackled this challenge within the EHR domain. Our work addresses this gap by adapting computer vision (CV) algorithms to mitigate the impact of label noise in DL models trained on EHR data. Notably, it remains uncertain whether CV methods, when applied to the EHR domain, will prove effective, given the substantial divergence between the two domains. We present empirical evidence demonstrating that these methods, whether used individually or in combination, can substantially enhance model performance when applied to EHR data, especially in the presence of noisy/incorrect labels. We validate our methods and underscore their practical utility in real-world EHR data, specifically in the context of COVID-19 diagnosis. Our study highlights the effectiveness of CV methods in the EHR domain, making a valuable contribution to the advancement of healthcare analytics and research.

Continuous patient state attention model for addressing irregularity in electronic health records.

BACKGROUND: Irregular time series (ITS) are common in healthcare as patient data is recorded in an electronic health record (EHR) system as per clinical guidelines/requirements but not for research and depends on a patient's health status. Due to irregularity, it is challenging to develop machine learning techniques to uncover vast intelligence hidden in EHR big data, without losing performance on downstream patient outcome prediction tasks. METHODS: In this paper, we propose Perceiver, a cross-attention-based transformer variant that is computationally efficient and can handle long sequences of time series in healthcare. We further develop continuous patient state attention models, using Perceiver and transformer to deal with ITS in EHR. The continuous patient state models utilise neural ordinary differential equations to learn patient health dynamics, i.e., patient health trajectory from observed irregular time steps, which enables them to sample patient state at any time. RESULTS: The proposed models' performance on in-hospital mortality prediction task on PhysioNet-2012 challenge and MIMIC-III datasets is examined. Perceiver model either outperforms or performs at par with baselines, and reduces computations by about nine times when compared to the transformer model, with no significant loss of performance. Experiments to examine irregularity in healthcare reveal that continuous patient state models outperform baselines. Moreover, the predictive uncertainty of the model is used to refer extremely uncertain cases to clinicians, which enhances the model's performance. Code is publicly available and verified at https://codeocean.com/capsule/4587224 . CONCLUSIONS: Perceiver presents a computationally efficient potential alternative for processing long sequences of time series in healthcare, and the continuous patient state attention models outperform the traditional and advanced techniques to handle irregularity in the time series. Moreover, the predictive uncertainty of the model helps in the development of transparent and trustworthy systems, which can be utilised as per the availability of clinicians.

Temporal dynamics unleashed: Elevating variational graph attention.

This research introduces the Variational Graph Attention Dynamics (VarGATDyn), addressing the complexities of dynamic graph representation learning, where existing models, tailored for static graphs, prove inadequate. VarGATDyn melds attention mechanisms with a Markovian assumption to surpass the challenges of maintaining temporal consistency and the extensive dataset requirements typical of RNN-based frameworks. It harnesses the strengths of the Variational Graph Auto-Encoder (VGAE) framework, Graph Attention Networks (GAT), and Gaussian Mixture Models (GMM) to adeptly navigate the temporal and structural intricacies of dynamic graphs. Through the strategic application of GMMs, the model handles multimodal patterns, thereby rectifying misalignments between prior and estimated posterior distributions. An innovative multiple-learning methodology bolsters the model's adaptability, leading to an encompassing and effective learning process. Empirical tests underscore VarGATDyn's dominance in dynamic link prediction across various datasets, highlighting its proficiency in capturing multimodal distributions and temporal dynamics.

An end-to-end heterogeneous graph attention network for Mycobacterium tuberculosis drug-resistance prediction.

Antimicrobial resistance (AMR) poses a threat to global public health. To mitigate the impacts of AMR, it is important to identify the molecular mechanisms of AMR and thereby determine optimal therapy as early as possible. Conventional machine learning-based drug-resistance analyses assume genetic variations to be homogeneous, thus not distinguishing between coding and intergenic sequences. In this study, we represent genetic data from Mycobacterium tuberculosis as a graph, and then adopt a deep graph learning method-heterogeneous graph attention network ('HGAT-AMR')-to predict anti-tuberculosis (TB) drug resistance. The HGAT-AMR model is able to accommodate incomplete phenotypic profiles, as well as provide 'attention scores' of genes and single nucleotide polymorphisms (SNPs) both at a population level and for individual samples. These scores encode the inputs, which the model is 'paying attention to' in making its drug resistance predictions. The results show that the proposed model generated the best area under the receiver operating characteristic (AUROC) for isoniazid and rifampicin (98.53 and 99.10%), the best sensitivity for three first-line drugs (94.91% for isoniazid, 96.60% for ethambutol and 90.63% for pyrazinamide), and maintained performance when the data were associated with incomplete phenotypes (i.e. for those isolates for which phenotypic data for some drugs were missing). We also demonstrate that the model successfully identifies genes and SNPs associated with drug resistance, mitigating the impact of resistance profile while considering particular drug resistance, which is consistent with domain knowledge.

A Stacked Long Short-Term Memory Approach for Predictive Blood Glucose Monitoring in Women with Gestational Diabetes Mellitus.

Gestational diabetes mellitus (GDM) is a subtype of diabetes that develops during pregnancy. Managing blood glucose (BG) within the healthy physiological range can reduce clinical complications for women with gestational diabetes. The objectives of this study are to (1) develop benchmark glucose prediction models with long short-term memory (LSTM) recurrent neural network models using time-series data collected from the GDm-Health platform, (2) compare the prediction accuracy with published results, and (3) suggest an optimized clinical review schedule with the potential to reduce the overall number of blood tests for mothers with stable and within-range glucose measurements. A total of 190,396 BG readings from 1110 patients were used for model development, validation and testing under three different prediction schemes: 7 days of BG readings to predict the next 7 or 14 days and 14 days to predict 14 days. Our results show that the optimized BG schedule based on a 7-day observational window to predict the BG of the next 14 days achieved the accuracies of the root mean square error (RMSE) = 0.958 ± 0.007, 0.876 ± 0.003, 0.898 ± 0.003, 0.622 ± 0.003, 0.814 ± 0.009 and 0.845 ± 0.005 for the after-breakfast, after-lunch, after-dinner, before-breakfast, before-lunch and before-dinner predictions, respectively. This is the first machine learning study that suggested an optimized blood glucose monitoring frequency, which is 7 days to monitor the next 14 days based on the accuracy of blood glucose prediction. Moreover, the accuracy of our proposed model based on the fingerstick blood glucose test is on par with the prediction accuracies compared with the benchmark performance of one-hour prediction models using continuous glucose monitoring (CGM) readings. In conclusion, the stacked LSTM model is a promising approach for capturing the patterns in time-series data, resulting in accurate predictions of BG levels. Using a deep learning model with routine fingerstick glucose collection is a promising, predictable and low-cost solution for BG monitoring for women with gestational diabetes.

Improving Diagnostics with Deep Forest Applied to Electronic Health Records.

An electronic health record (EHR) is a vital high-dimensional part of medical concepts. Discovering implicit correlations in the information of this data set and the research and informative aspects can improve the treatment and management process. The challenge of concern is the data sources' limitations in finding a stable model to relate medical concepts and use these existing connections. This paper presents Patient Forest, a novel end-to-end approach for learning patient representations from tree-structured data for readmission and mortality prediction tasks. By leveraging statistical features, the proposed model is able to provide an accurate and reliable classifier for predicting readmission and mortality. Experiments on MIMIC-III and eICU datasets demonstrate Patient Forest outperforms existing machine learning models, especially when the training data are limited. Additionally, a qualitative evaluation of Patient Forest is conducted by visualising the learnt representations in 2D space using the t-SNE, which further confirms the effectiveness of the proposed model in learning EHR representations.

2D-WinSpatt-Net: A Dual Spatial Self-Attention Vision Transformer Boosts Classification of Tetanus Severity for Patients Wearing ECG Sensors in Low- and Middle-Income Countries.

Tetanus is a life-threatening bacterial infection that is often prevalent in low- and middle-income countries (LMIC), Vietnam included. Tetanus affects the nervous system, leading to muscle stiffness and spasms. Moreover, severe tetanus is associated with autonomic nervous system (ANS) dysfunction. To ensure early detection and effective management of ANS dysfunction, patients require continuous monitoring of vital signs using bedside monitors. Wearable electrocardiogram (ECG) sensors offer a more cost-effective and user-friendly alternative to bedside monitors. Machine learning-based ECG analysis can be a valuable resource for classifying tetanus severity; however, using existing ECG signal analysis is excessively time-consuming. Due to the fixed-sized kernel filters used in traditional convolutional neural networks (CNNs), they are limited in their ability to capture global context information. In this work, we propose a 2D-WinSpatt-Net, which is a novel Vision Transformer that contains both local spatial window self-attention and global spatial self-attention mechanisms. The 2D-WinSpatt-Net boosts the classification of tetanus severity in intensive-care settings for LMIC using wearable ECG sensors. The time series imaging-continuous wavelet transforms-is transformed from a one-dimensional ECG signal and input to the proposed 2D-WinSpatt-Net. In the classification of tetanus severity levels, 2D-WinSpatt-Net surpasses state-of-the-art methods in terms of performance and accuracy. It achieves remarkable results with an F1 score of 0.88 ± 0.00, precision of 0.92 ± 0.02, recall of 0.85 ± 0.01, specificity of 0.96 ± 0.01, accuracy of 0.93 ± 0.02 and AUC of 0.90 ± 0.00.

Omicron-associated changes in SARS-CoV-2 symptoms in the United Kingdom.

BACKGROUND: The SARS-CoV-2 Delta variant has been replaced by the highly transmissible Omicron BA.1 variant, and subsequently by Omicron BA.2. It is important to understand how these changes in dominant variants affect reported symptoms, while also accounting for symptoms arising from other co-circulating respiratory viruses. METHODS: In a nationally representative UK community study, the COVID-19 Infection Survey, we investigated symptoms in PCR-positive infection episodes vs. PCR-negative study visits over calendar time, by age and vaccination status, comparing periods when the Delta, Omicron BA.1 and BA.2 variants were dominant. RESULTS: Between October-2020 and April-2022, 120,995 SARS-CoV-2 PCR-positive episodes occurred in 115,886 participants, with 70,683 (58%) reporting symptoms. The comparator comprised 4,766,366 PCR-negative study visits (483,894 participants); 203,422 (4%) reporting symptoms. Symptom reporting in PCR-positives varied over time, with a marked reduction in loss of taste/smell as Omicron BA.1 dominated, maintained with BA.2 (44%/45% 17 October 2021, 16%/13% 2 January 2022, 15%/12% 27 March 2022). Cough, fever, shortness of breath, myalgia, fatigue/weakness and headache also decreased after Omicron BA.1 dominated, but sore throat increased, the latter to a greater degree than concurrent increases in PCR-negatives. Fatigue/weakness increased again after BA.2 dominated, although to a similar degree to concurrent increases in PCR-negatives. Symptoms were consistently more common in adults aged 18-65 years than in children or older adults. CONCLUSIONS: Increases in sore throat (also common in the general community), and a marked reduction in loss of taste/smell, make Omicron harder to detect with symptom-based testing algorithms, with implications for institutional and national testing policies.

Global SARS-CoV-2 seroprevalence from January 2020 to April 2022: A systematic review and meta-analysis of standardized population-based studies.

BACKGROUND: Our understanding of the global scale of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection remains incomplete: Routine surveillance data underestimate infection and cannot infer on population immunity; there is a predominance of asymptomatic infections, and uneven access to diagnostics. We meta-analyzed SARS-CoV-2 seroprevalence studies, standardized to those described in the World Health Organization's Unity protocol (WHO Unity) for general population seroepidemiological studies, to estimate the extent of population infection and seropositivity to the virus 2 years into the pandemic. METHODS AND FINDINGS: We conducted a systematic review and meta-analysis, searching MEDLINE, Embase, Web of Science, preprints, and grey literature for SARS-CoV-2 seroprevalence published between January 1, 2020 and May 20, 2022. The review protocol is registered with PROSPERO (CRD42020183634). We included general population cross-sectional and cohort studies meeting an assay quality threshold (90% sensitivity, 97% specificity; exceptions for humanitarian settings). We excluded studies with an unclear or closed population sample frame. Eligible studies-those aligned with the WHO Unity protocol-were extracted and critically appraised in duplicate, with risk of bias evaluated using a modified Joanna Briggs Institute checklist. We meta-analyzed seroprevalence by country and month, pooling to estimate regional and global seroprevalence over time; compared seroprevalence from infection to confirmed cases to estimate underascertainment; meta-analyzed differences in seroprevalence between demographic subgroups such as age and sex; and identified national factors associated with seroprevalence using meta-regression. We identified 513 full texts reporting 965 distinct seroprevalence studies (41% low- and middle-income countries [LMICs]) sampling 5,346,069 participants between January 2020 and April 2022, including 459 low/moderate risk of bias studies with national/subnational scope in further analysis. By September 2021, global SARS-CoV-2 seroprevalence from infection or vaccination was 59.2%, 95% CI [56.1% to 62.2%]. Overall seroprevalence rose steeply in 2021 due to infection in some regions (e.g., 26.6% [24.6 to 28.8] to 86.7% [84.6% to 88.5%] in Africa in December 2021) and vaccination and infection in others (e.g., 9.6% [8.3% to 11.0%] in June 2020 to 95.9% [92.6% to 97.8%] in December 2021, in European high-income countries [HICs]). After the emergence of Omicron in March 2022, infection-induced seroprevalence rose to 47.9% [41.0% to 54.9%] in Europe HIC and 33.7% [31.6% to 36.0%] in Americas HIC. In 2021 Quarter Three (July to September), median seroprevalence to cumulative incidence ratios ranged from around 2:1 in the Americas and Europe HICs to over 100:1 in Africa (LMICs). Children 0 to 9 years and adults 60+ were at lower risk of seropositivity than adults 20 to 29 (p < 0.001 and p = 0.005, respectively). In a multivariable model using prevaccination data, stringent public health and social measures were associated with lower seroprevalence (p = 0.02). The main limitations of our methodology include that some estimates were driven by certain countries or populations being overrepresented. CONCLUSIONS: In this study, we observed that global seroprevalence has risen considerably over time and with regional variation; however, over one-third of the global population are seronegative to the SARS-CoV-2 virus. Our estimates of infections based on seroprevalence far exceed reported Coronavirus Disease 2019 (COVID-19) cases. Quality and standardized seroprevalence studies are essential to inform COVID-19 response, particularly in resource-limited regions.

Detecting Deteriorating Patients in the Hospital: Development and Validation of a Novel Scoring System.

Rationale: Late recognition of patient deterioration in hospital is associated with worse outcomes, including higher mortality. Despite the widespread introduction of early warning score (EWS) systems and electronic health records, deterioration still goes unrecognized. Objectives: To develop and externally validate a Hospital- wide Alerting via Electronic Noticeboard (HAVEN) system to identify hospitalized patients at risk of reversible deterioration. Methods: This was a retrospective cohort study of patients 16 years of age or above admitted to four UK hospitals. The primary outcome was cardiac arrest or unplanned admission to the ICU. We used patient data (vital signs, laboratory tests, comorbidities, and frailty) from one hospital to train a machine-learning model (gradient boosting trees). We internally and externally validated the model and compared its performance with existing scoring systems (including the National EWS, laboratory-based acute physiology score, and electronic cardiac arrest risk triage score). Measurements and Main Results: We developed the HAVEN model using 230,415 patient admissions to a single hospital. We validated HAVEN on 266,295 admissions to four hospitals. HAVEN showed substantially higher discrimination (c-statistic, 0.901 [95% confidence interval, 0.898-0.903]) for the primary outcome within 24 hours of each measurement than other published scoring systems (which range from 0.700 [0.696-0.704] to 0.863 [0.860-0.865]). With a precision of 10%, HAVEN was able to identify 42% of cardiac arrests or unplanned ICU admissions with a lead time of up to 48 hours in advance, compared with 22% by the next best system. Conclusions: The HAVEN machine-learning algorithm for early identification of in-hospital deterioration significantly outperforms other published scores such as the National EWS.

A Deep Learning Approach for the Assessment of Signal Quality of Non-Invasive Foetal Electrocardiography.

Non-invasive foetal electrocardiography (NI-FECG) has become an important prenatal monitoring method in the hospital. However, due to its susceptibility to non-stationary noise sources and lack of robust extraction methods, the capture of high-quality NI-FECG remains a challenge. Recording waveforms of sufficient quality for clinical use typically requires human visual inspection of each recording. A Signal Quality Index (SQI) can help to automate this task but, contrary to adult ECG, work on SQIs for NI-FECG is sparse. In this paper, a multi-channel signal quality classifier for NI-FECG waveforms is presented. The model can be used during the capture of NI-FECG to assist technicians to record high-quality waveforms, which is currently a labour-intensive task. A Convolutional Neural Network (CNN) is trained to distinguish between NI-FECG segments of high and low quality. NI-FECG recordings with one maternal channel and three abdominal channels were collected from 100 subjects during a routine hospital screening (102.6 min of data). The model achieves an average 10-fold cross-validated AUC of 0.95 ± 0.02. The results show that the model can reliably assess the FECG signal quality on our dataset. The proposed model can improve the automated capture and analysis of NI-FECG as well as reduce technician labour time.

Sepsis Mortality Prediction Using Wearable Monitoring in Low-Middle Income Countries.

Sepsis is associated with high mortality-particularly in low-middle income countries (LMICs). Critical care management of sepsis is challenging in LMICs due to the lack of care providers and the high cost of bedside monitors. Recent advances in wearable sensor technology and machine learning (ML) models in healthcare promise to deliver new ways of digital monitoring integrated with automated decision systems to reduce the mortality risk in sepsis. In this study, firstly, we aim to assess the feasibility of using wearable sensors instead of traditional bedside monitors in the sepsis care management of hospital admitted patients, and secondly, to introduce automated prediction models for the mortality prediction of sepsis patients. To this end, we continuously monitored 50 sepsis patients for nearly 24 h after their admission to the Hospital for Tropical Diseases in Vietnam. We then compared the performance and interpretability of state-of-the-art ML models for the task of mortality prediction of sepsis using the heart rate variability (HRV) signal from wearable sensors and vital signs from bedside monitors. Our results show that all ML models trained on wearable data outperformed ML models trained on data gathered from the bedside monitors for the task of mortality prediction with the highest performance (area under the precision recall curve = 0.83) achieved using time-varying features of HRV and recurrent neural networks. Our results demonstrate that the integration of automated ML prediction models with wearable technology is well suited for helping clinicians who manage sepsis patients in LMICs to reduce the mortality risk of sepsis.

Classification of Tetanus Severity in Intensive-Care Settings for Low-Income Countries Using Wearable Sensing.

Infectious diseases remain a common problem in low- and middle-income countries, including in Vietnam. Tetanus is a severe infectious disease characterized by muscle spasms and complicated by autonomic nervous system dysfunction in severe cases. Patients require careful monitoring using electrocardiograms (ECGs) to detect deterioration and the onset of autonomic nervous system dysfunction as early as possible. Machine learning analysis of ECG has been shown of extra value in predicting tetanus severity, however any additional ECG signal analysis places a high demand on time-limited hospital staff and requires specialist equipment. Therefore, we present a novel approach to tetanus monitoring from low-cost wearable sensors combined with a deep-learning-based automatic severity detection. This approach can automatically triage tetanus patients and reduce the burden on hospital staff. In this study, we propose a two-dimensional (2D) convolutional neural network with a channel-wise attention mechanism for the binary classification of ECG signals. According to the Ablett classification of tetanus severity, we define grades 1 and 2 as mild tetanus and grades 3 and 4 as severe tetanus. The one-dimensional ECG time series signals are transformed into 2D spectrograms. The 2D attention-based network is designed to extract the features from the input spectrograms. Experiments demonstrate a promising performance for the proposed method in tetanus classification with an F1 score of 0.79 ± 0.03, precision of 0.78 ± 0.08, recall of 0.82 ± 0.05, specificity of 0.85 ± 0.08, accuracy of 0.84 ± 0.04 and AUC of 0.84 ± 0.03.

Prediction of Susceptibility to First-Line Tuberculosis Drugs by DNA Sequencing.

BACKGROUND: The World Health Organization recommends drug-susceptibility testing of Mycobacterium tuberculosis complex for all patients with tuberculosis to guide treatment decisions and improve outcomes. Whether DNA sequencing can be used to accurately predict profiles of susceptibility to first-line antituberculosis drugs has not been clear. METHODS: We obtained whole-genome sequences and associated phenotypes of resistance or susceptibility to the first-line antituberculosis drugs isoniazid, rifampin, ethambutol, and pyrazinamide for isolates from 16 countries across six continents. For each isolate, mutations associated with drug resistance and drug susceptibility were identified across nine genes, and individual phenotypes were predicted unless mutations of unknown association were also present. To identify how whole-genome sequencing might direct first-line drug therapy, complete susceptibility profiles were predicted. These profiles were predicted to be susceptible to all four drugs (i.e., pansusceptible) if they were predicted to be susceptible to isoniazid and to the other drugs or if they contained mutations of unknown association in genes that affect susceptibility to the other drugs. We simulated the way in which the negative predictive value changed with the prevalence of drug resistance. RESULTS: A total of 10,209 isolates were analyzed. The largest proportion of phenotypes was predicted for rifampin (9660 [95.4%] of 10,130) and the smallest was predicted for ethambutol (8794 [89.8%] of 9794). Resistance to isoniazid, rifampin, ethambutol, and pyrazinamide was correctly predicted with 97.1%, 97.5%, 94.6%, and 91.3% sensitivity, respectively, and susceptibility to these drugs was correctly predicted with 99.0%, 98.8%, 93.6%, and 96.8% specificity. Of the 7516 isolates with complete phenotypic drug-susceptibility profiles, 5865 (78.0%) had complete genotypic predictions, among which 5250 profiles (89.5%) were correctly predicted. Among the 4037 phenotypic profiles that were predicted to be pansusceptible, 3952 (97.9%) were correctly predicted. CONCLUSIONS: Genotypic predictions of the susceptibility of M. tuberculosis to first-line drugs were found to be correlated with phenotypic susceptibility to these drugs. (Funded by the Bill and Melinda Gates Foundation and others.).

SDNN24 Estimation from Semi-Continuous HR Measures.

The standard deviation of the interval between QRS complexes recorded over 24 h (SDNN24) is an important metric of cardiovascular health. Wrist-worn fitness wearable devices record heart beats 24/7 having a complete overview of users' heart status. Due to motion artefacts affecting QRS complexes recording, and the different nature of the heart rate sensor used on wearable devices compared to ECG, traditionally used to compute SDNN24, the estimation of this important Heart Rate Variability (HRV) metric has never been performed from wearable data. We propose an innovative approach to estimate SDNN24 only exploiting the Heart Rate (HR) that is normally available on wearable fitness trackers and less affected by data noise. The standard deviation of inter-beats intervals (SDNN24) and the standard deviation of the Average inter-beats intervals (ANN) derived from the HR (obtained in a time window with defined duration, i.e., 1, 5, 10, 30 and 60 min), i.e., ANN=60HR (SDANNHR24), were calculated over 24 h. Power spectrum analysis using the Lomb-Scargle Peridogram was performed to assess frequency domain HRV parameters (Ultra Low Frequency, Very Low Frequency, Low Frequency, and High Frequency). Due to the fact that SDNN24 reflects the total power of the power of the HRV spectrum, the values estimated from HR measures (SDANNHR24) underestimate the real values because of the high frequencies that are missing. Subjects with low and high cardiovascular risk show different power spectra. In particular, differences are detected in Ultra Low and Very Low frequencies, while similar results are shown in Low and High frequencies. For this reason, we found that HR measures contain enough information to discriminate cardiovascular risk. Semi-continuous measures of HR throughout 24 h, as measured by most wrist-worn fitness wearable devices, should be sufficient to estimate SDNN24 and cardiovascular risk.

DeepAMR for predicting co-occurrent resistance of Mycobacterium tuberculosis.

MOTIVATION: Resistance co-occurrence within first-line anti-tuberculosis (TB) drugs is a common phenomenon. Existing methods based on genetic data analysis of Mycobacterium tuberculosis (MTB) have been able to predict resistance of MTB to individual drugs, but have not considered the resistance co-occurrence and cannot capture latent structure of genomic data that corresponds to lineages. RESULTS: We used a large cohort of TB patients from 16 countries across six continents where whole-genome sequences for each isolate and associated phenotype to anti-TB drugs were obtained using drug susceptibility testing recommended by the World Health Organization. We then proposed an end-to-end multi-task model with deep denoising auto-encoder (DeepAMR) for multiple drug classification and developed DeepAMR_cluster, a clustering variant based on DeepAMR, for learning clusters in latent space of the data. The results showed that DeepAMR outperformed baseline model and four machine learning models with mean AUROC from 94.4% to 98.7% for predicting resistance to four first-line drugs [i.e. isoniazid (INH), ethambutol (EMB), rifampicin (RIF), pyrazinamide (PZA)], multi-drug resistant TB (MDR-TB) and pan-susceptible TB (PANS-TB: MTB that is susceptible to all four first-line anti-TB drugs). In the case of INH, EMB, PZA and MDR-TB, DeepAMR achieved its best mean sensitivity of 94.3%, 91.5%, 87.3% and 96.3%, respectively. While in the case of RIF and PANS-TB, it generated 94.2% and 92.2% sensitivity, which were lower than baseline model by 0.7% and 1.9%, respectively. t-SNE visualization shows that DeepAMR_cluster captures lineage-related clusters in the latent space. AVAILABILITY AND IMPLEMENTATION: The details of source code are provided at http://www.robots.ox.ac.uk/∼davidc/code.php. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Application of machine learning techniques to tuberculosis drug resistance analysis.

MOTIVATION: Timely identification of Mycobacterium tuberculosis (MTB) resistance to existing drugs is vital to decrease mortality and prevent the amplification of existing antibiotic resistance. Machine learning methods have been widely applied for timely predicting resistance of MTB given a specific drug and identifying resistance markers. However, they have been not validated on a large cohort of MTB samples from multi-centers across the world in terms of resistance prediction and resistance marker identification. Several machine learning classifiers and linear dimension reduction techniques were developed and compared for a cohort of 13 402 isolates collected from 16 countries across 6 continents and tested 11 drugs. RESULTS: Compared to conventional molecular diagnostic test, area under curve of the best machine learning classifier increased for all drugs especially by 23.11%, 15.22% and 10.14% for pyrazinamide, ciprofloxacin and ofloxacin, respectively (P < 0.01). Logistic regression and gradient tree boosting found to perform better than other techniques. Moreover, logistic regression/gradient tree boosting with a sparse principal component analysis/non-negative matrix factorization step compared with the classifier alone enhanced the best performance in terms of F1-score by 12.54%, 4.61%, 7.45% and 9.58% for amikacin, moxifloxacin, ofloxacin and capreomycin, respectively, as well increasing area under curve for amikacin and capreomycin. Results provided a comprehensive comparison of various techniques and confirmed the application of machine learning for better prediction of the large diverse tuberculosis data. Furthermore, mutation ranking showed the possibility of finding new resistance/susceptible markers. AVAILABILITY AND IMPLEMENTATION: The source code can be found at http://www.robots.ox.ac.uk/ davidc/code.php. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Error Estimation of Ultra-Short Heart Rate Variability Parameters: Effect of Missing Data Caused by Motion Artifacts.

Application of ultra-short Heart Rate Variability (HRV) is desirable in order to increase the applicability of HRV features to wrist-worn wearable devices equipped with heart rate sensors that are nowadays becoming more and more popular in people's daily life. This study is focused in particular on the the two most used HRV parameters, i.e., the standard deviation of inter-beat intervals (SDNN) and the root Mean Squared error of successive inter-beat intervals differences (rMSSD). The huge problem of extracting these HRV parameters from wrist-worn devices is that their data are affected by the motion artifacts. For this reason, estimating the error caused by this huge quantity of missing values is fundamental to obtain reliable HRV parameters from these devices. To this aim, we simulate missing values induced by motion artifacts (from 0 to 70%) in an ultra-short time window (i.e., from 4 min to 30 s) by the random walk Gilbert burst model in 22 young healthy subjects. In addition, 30 s and 2 min ultra-short time windows are required to estimate rMSSD and SDNN, respectively. Moreover, due to the fact that ultra-short time window does not permit assessing very low frequencies, and the SDNN is highly affected by these frequencies, the bias for estimating SDNN continues to increase as the time window length decreases. On the contrary, a small error is detected in rMSSD up to 30 s due to the fact that it is highly affected by high frequencies which are possible to be evaluated even if the time window length decreases. Finally, the missing values have a small effect on rMSSD and SDNN estimation. As a matter of fact, the HRV parameter errors increase slightly as the percentage of missing values increase.