RESUMO
INTRODUCTION: Machine learning algorithms are expected to work side-by-side with humans in decision-making pipelines. Thus, the ability of classifiers to make reliable decisions is of paramount importance. Deep neural networks (DNNs) represent the state-of-the-art models to address real-world classification. Although the strength of activation in DNNs is often correlated with the network's confidence, in-depth analyses are needed to establish whether they are well calibrated. METHOD: In this paper, we demonstrate the use of DNN-based classification tools to benefit cancer registries by automating information extraction of disease at diagnosis and at surgery from electronic text pathology reports from the US National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) population-based cancer registries. In particular, we introduce multiple methods for selective classification to achieve a target level of accuracy on multiple classification tasks while minimizing the rejection amount-that is, the number of electronic pathology reports for which the model's predictions are unreliable. We evaluate the proposed methods by comparing our approach with the current in-house deep learning-based abstaining classifier. RESULTS: Overall, all the proposed selective classification methods effectively allow for achieving the targeted level of accuracy or higher in a trade-off analysis aimed to minimize the rejection rate. On in-distribution validation and holdout test data, with all the proposed methods, we achieve on all tasks the required target level of accuracy with a lower rejection rate than the deep abstaining classifier (DAC). Interpreting the results for the out-of-distribution test data is more complex; nevertheless, in this case as well, the rejection rate from the best among the proposed methods achieving 97% accuracy or higher is lower than the rejection rate based on the DAC. CONCLUSIONS: We show that although both approaches can flag those samples that should be manually reviewed and labeled by human annotators, the newly proposed methods retain a larger fraction and do so without retraining-thus offering a reduced computational cost compared with the in-house deep learning-based abstaining classifier.
Assuntos
Aprendizado Profundo , Humanos , Incerteza , Redes Neurais de Computação , Algoritmos , Aprendizado de MáquinaRESUMO
In the opening months of the pandemic, the need for situational awareness was urgent. Forecasting models such as the Susceptible-Infectious-Recovered (SIR) model were hampered by limited testing data and key information on mobility, contact tracing, and local policy variations would not be consistently available for months. New case counts from sources like John Hopkins University and the NY Times were systematically reliable. Using these data, we developed the novel COVID County Situational Awareness Tool (CCSAT) for reliable monitoring and decision support. In CCSAT, we developed a retrospective seven-day moving window semantic map of county-level disease magnitude and acceleration that smoothed noisy daily variations. We also developed a novel Bayesian model that reliably forecasted county-level magnitude and acceleration for the upcoming week based on population and new case count data. Together these formed a robust operational update including county-level maps of new case rate changes, estimates of new cases in the upcoming week, and measures of model reliability. We found CCSAT provided stable, reliable estimates across the seven-day time window, with the greatest errors occurring in cases of anomalous, single day spikes. In this paper, we provide CCSAT details and apply it to a single week in June 2020.
RESUMO
BACKGROUND: The search for statistically significant relationships between molecular markers and outcomes is challenging when dealing with high-dimensional, noisy and collinear multivariate omics data, such as metabolomic profiles. Permutation procedures allow for the estimation of adjusted significance levels without assuming independence among metabolomic variables. Nevertheless, the complex non-normal structure of metabolic profiles and outcomes may bias the permutation results leading to overly conservative threshold estimates i.e. lower than those from a Bonferroni or Sidak correction. METHODS: Within a univariate permutation procedure we employ parametric simulation methods based on the multivariate (log-)Normal distribution to obtain adjusted significance levels which are consistent across different outcomes while effectively controlling the type I error rate. Next, we derive an alternative closed-form expression for the estimation of the number of non-redundant metabolic variates based on the spectral decomposition of their correlation matrix. The performance of the method is tested for different model parametrizations and across a wide range of correlation levels of the variates using synthetic and real data sets. RESULTS: Both the permutation-based formulation and the more practical closed form expression are found to give an effective indication of the number of independent metabolic effects exhibited by the system, while guaranteeing that the derived adjusted threshold is stable across outcome measures with diverse properties.
Assuntos
Metaboloma , Metabolômica , Modelos Biológicos , Marcadores Genéticos/genética , Metabolômica/métodos , Distribuições EstatísticasRESUMO
Physical environment plays a key role in determining human health risks. Exposure to toxins, weather extremes, degraded air and water quality, high levels of noise and limited accessibility to green areas can negatively affect health. Furthermore, adverse environmental exposures are often correlated with each other and with socioeconomic status, thereby compounding disadvantages in marginalized populations. Moreover, despite their importance in determining human health risks, the role of multiple environmental exposures is not well studied, and only a few resources contain aggregate environmental exposure data and only for selected areas of the contiguous US. To fill these gaps, we took a cumulative approach to measuring the environment by generating a composite Multi-Exposure Environmental Index (MEEI) as a US Census Tract-level summary of key environmental factors with known health effects. This measure quantifies multiple environmental exposures in the same area that can result in additive and synergistic effects on health outcomes. This information is crucial to better understand and possibly leverage environmental determinants of health for informed policy-making and intervention.
RESUMO
CONTEXT: Humans respond profoundly to changes in diet, while nutrition and environment have a great impact on population health. It is therefore important to deeply characterize the human nutritional responses. OBJECTIVE: Endocrine parameters and the metabolome of human plasma are rapidly responding to acute nutritional interventions such as caloric restriction or a glucose challenge. It is less well understood whether the plasma proteome would be equally dynamic, and whether it could be a source of corresponding biomarkers. METHODS: We used high-throughput mass spectrometry to determine changes in the plasma proteome of i) 10 healthy, young, male individuals in response to 2 days of acute caloric restriction followed by refeeding; ii) 200 individuals of the Ely epidemiological study before and after a glucose tolerance test at 4 time points (0, 30, 60, 120 minutes); and iii) 200 random individuals from the Generation Scotland study. We compared the proteomic changes detected with metabolome data and endocrine parameters. RESULTS: Both caloric restriction and the glucose challenge substantially impacted the plasma proteome. Proteins responded across individuals or in an individual-specific manner. We identified nutrient-responsive plasma proteins that correlate with changes in the metabolome, as well as with endocrine parameters. In particular, our study highlights the role of apolipoprotein C1 (APOC1), a small, understudied apolipoprotein that was affected by caloric restriction and dominated the response to glucose consumption and differed in abundance between individuals with and without type 2 diabetes. CONCLUSION: Our study identifies APOC1 as a dominant nutritional responder in humans and highlights the interdependency of acute nutritional response proteins and the endocrine system.
Assuntos
Diabetes Mellitus Tipo 2 , Proteoma , Humanos , Masculino , Proteômica , Glucose , Restrição CalóricaRESUMO
Mortality prediction for intensive care unit (ICU) patients is crucial for improving outcomes and efficient utilization of resources. Accessibility of electronic health records (EHR) has enabled data-driven predictive modeling using machine learning. However, very few studies rely solely on unstructured clinical notes from the EHR for mortality prediction. In this work, we propose a framework to predict short, mid, and long-term mortality in adult ICU patients using unstructured clinical notes from the MIMIC III database, natural language processing (NLP), and machine learning (ML) models. Depending on the statistical description of the patients' length of stay, we define the short-term as 48-hour and 4-day period, the mid-term as 7-day and 10-day period, and the long-term as 15-day and 30-day period after admission. We found that by only using clinical notes within the 24 hours of admission, our framework can achieve a high area under the receiver operating characteristics (AU-ROC) score for short, mid and long-term mortality prediction tasks. The test AU-ROC scores are 0.87, 0.83, 0.83, 0.82, 0.82, and 0.82 for 48-hour, 4-day, 7-day, 10-day, 15-day, and 30-day period mortality prediction, respectively. We also provide a comparative study among three types of feature extraction techniques from NLP: frequency-based technique, fixed embedding-based technique, and dynamic embedding-based technique. Lastly, we provide an interpretation of the NLP-based predictive models using feature-importance scores.
Assuntos
Mortalidade Hospitalar , Aprendizado de Máquina , Área Sob a Curva , Bases de Dados Factuais , Registros Eletrônicos de Saúde , Humanos , Unidades de Terapia Intensiva , Tempo de Internação , Modelos Logísticos , Curva ROCRESUMO
Objectives: The International Classification of Childhood Cancer (ICCC) facilitates the effective classification of a heterogeneous group of cancers in the important pediatric population. However, there has been no development of machine learning models for the ICCC classification. We developed deep learning-based information extraction models from cancer pathology reports based on the ICD-O-3 coding standard. In this article, we describe extending the models to perform ICCC classification. Materials and Methods: We developed 2 models, ICD-O-3 classification and ICCC recoding (Model 1) and direct ICCC classification (Model 2), and 4 scenarios subject to the training sample size. We evaluated these models with a corpus consisting of 29â206 reports with age at diagnosis between 0 and 19 from 6 state cancer registries. Results: Our findings suggest that the direct ICCC classification (Model 2) is substantially better than reusing the ICD-O-3 classification model (Model 1). Applying the uncertainty quantification mechanism to assess the confidence of the algorithm in assigning a code demonstrated that the model achieved a micro-F1 score of 0.987 while abstaining (not sufficiently confident to assign a code) on only 14.8% of ambiguous pathology reports. Conclusions: Our experimental results suggest that the machine learning-based automatic information extraction from childhood cancer pathology reports in the ICCC is a reliable means of supplementing human annotators at state cancer registries by reading and abstracting the majority of the childhood cancer pathology reports accurately and reliably.
RESUMO
Mass spectrometry technologies are widely used in the fields of ionomics and metabolomics to simultaneously profile the intracellular concentrations of, e.g., amino acids or elements in genome-wide mutant libraries. These molecular or sub-molecular features are generally non-Gaussian and their covariance reveals patterns of correlations that reflect the system nature of the cell biochemistry and biology. Here, we introduce two similarity measures, the Mahalanobis cosine and the hybrid Mahalanobis cosine, that enforce information from the empirical covariance matrix of omics data from high-throughput screening and that can be used to quantify similarities between the profiled features of different mutants. We evaluate the performance of these similarity measures in the task of inferring and integrating genetic networks from short-profile ionomics/metabolomics data through an analysis of experimental data sets related to the ionome and the metabolome of the model organism S. cerevisiae. The study of the resulting ionome-metabolome Saccharomyces cerevisiae multilayer genetic network, which encodes multiple omic-specific levels of correlations between genes, shows that the proposed measures can provide an alternative description of relations between biological processes when compared to the commonly used Pearson's correlation coefficient and have the potential to guide the construction of novel hypotheses on the function of uncharacterised genes.
RESUMO
BACKGROUND: Metabolomics is the comprehensive study of a multitude of small molecules to gain insight into an organism's metabolism. The research field is dynamic and expanding with applications across biomedical, biotechnological, and many other applied biological domains. Its computationally intensive nature has driven requirements for open data formats, data repositories, and data analysis tools. However, the rapid progress has resulted in a mosaic of independent, and sometimes incompatible, analysis methods that are difficult to connect into a useful and complete data analysis solution. FINDINGS: PhenoMeNal (Phenome and Metabolome aNalysis) is an advanced and complete solution to set up Infrastructure-as-a-Service (IaaS) that brings workflow-oriented, interoperable metabolomics data analysis platforms into the cloud. PhenoMeNal seamlessly integrates a wide array of existing open-source tools that are tested and packaged as Docker containers through the project's continuous integration process and deployed based on a kubernetes orchestration framework. It also provides a number of standardized, automated, and published analysis workflows in the user interfaces Galaxy, Jupyter, Luigi, and Pachyderm. CONCLUSIONS: PhenoMeNal constitutes a keystone solution in cloud e-infrastructures available for metabolomics. PhenoMeNal is a unique and complete solution for setting up cloud e-infrastructures through easy-to-use web interfaces that can be scaled to any custom public and private cloud environment. By harmonizing and automating software installation and configuration and through ready-to-use scientific workflow user interfaces, PhenoMeNal has succeeded in providing scientists with workflow-driven, reproducible, and shareable metabolomics data analysis platforms that are interfaced through standard data formats, representative datasets, versioned, and have been tested for reproducibility and interoperability. The elastic implementation of PhenoMeNal further allows easy adaptation of the infrastructure to other application areas and 'omics research domains.