Dataset Information

A modular approach to integrating multiple data sources into real-time clinical prediction for pediatric diarrhea.

ABSTRACT: Traditional clinical prediction models focus on parameters of the individual patient. For infectious diseases, sources external to the patient, including characteristics of prior patients and seasonal factors, may improve predictive performance. We describe the development of a predictive model that integrates multiple sources of data in a principled statistical framework using a post-test odds formulation. Our method enables electronic real-time updating and flexibility, such that components can be included or excluded according to data availability. We apply this method to the prediction of etiology of pediatric diarrhea, where 'pre-test' epidemiologic data may be highly informative. Diarrhea has a high burden in low-resource settings, and antibiotics are often over-prescribed. We demonstrate that our integrative method outperforms traditional prediction in accurately identifying cases with a viral etiology, and show that its clinical application, especially when used with an additional diagnostic test, could result in a 61% reduction in inappropriately prescribed antibiotics.

SUBMITTER: Brintz BJ

PROVIDER: S-EPMC7853717 | biostudies-literature | 2021 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A modular approach to integrating multiple data sources into real-time clinical prediction for pediatric diarrhea.

Brintz Ben J BJ Haaland Benjamin B Howard Joel J Chao Dennis L DL Proctor Joshua L JL Khan Ashraful I AI Ahmed Sharia M SM Keegan Lindsay T LT Greene Tom T Keita Adama Mamby AM Kotloff Karen L KL Platts-Mills James A JA Nelson Eric J EJ Levine Adam C AC Pavia Andrew T AT Leung Daniel T DT

eLife 20210202

Traditional clinical prediction models focus on parameters of the individual patient. For infectious diseases, sources external to the patient, including characteristics of prior patients and seasonal factors, may improve predictive performance. We describe the development of a predictive model that integrates multiple sources of data in a principled statistical framework using a post-test odds formulation. Our method enables electronic real-time updating and flexibility, such that components ca ...[more]

PMID: 33527894

Similar Datasets

Project description:BackgroundProtein function determination is a key challenge in the post-genomic era. Experimental determination of protein functions is accurate, but time-consuming and resource-intensive. A cost-effective alternative is to use the known information about sequence, structure, and functional properties of genes and proteins to predict functions using statistical methods. In this paper, we describe the Multi-Source k-Nearest Neighbor (MS-kNN) algorithm for function prediction, which finds k-nearest neighbors of a query protein based on different types of similarity measures and predicts its function by weighted averaging of its neighbors' functions. Specifically, we used 3 data sources to calculate the similarity scores: sequence similarity, protein-protein interactions, and gene expressions.ResultsWe report the results in the context of 2011 Critical Assessment of Function Annotation (CAFA). Prior to CAFA submission deadline, we evaluated our algorithm on 1,302 human test proteins that were represented in all 3 data sources. Using only the sequence similarity information, MS-kNN had term-based Area Under the Curve (AUC) accuracy of Gene Ontology (GO) molecular function predictions of 0.728 when 7,412 human training proteins were used, and 0.819 when 35,622 training proteins from multiple eukaryotic and prokaryotic organisms were used. By aggregating predictions from all three sources, the AUC was further improved to 0.848. Similar result was observed on prediction of GO biological processes. Testing on 595 proteins that were annotated after the CAFA submission deadline showed that overall MS-kNN accuracy was higher than that of baseline algorithms Gotcha and BLAST, which were based solely on sequence similarity information. Since only 10 of the 595 proteins were represented by all 3 data sources, and 66 by two data sources, the difference between 3-source and one-source MS-kNN was rather small.ConclusionsBased on our results, we have several useful insights: (1) the k-nearest neighbor algorithm is an efficient and effective model for protein function prediction; (2) it is beneficial to transfer functions across a wide range of organisms; (3) it is helpful to integrate multiple sources of protein information.

Project description:Reducing colonization of poultry flocks by Campylobacter spp. is a key strategy in the control and prevention human campylobacteriosis. Horizontal transmission of campylobacters, from in and around the farm, is the presumed route of flock colonization. However, the identification and prioritization of sources are confounded by the ubiquitous nature of these organisms in the environment, their poor rates of recovery by standard culture methods, and the need for cost-effective and timely methods for strain-specific comparison. A real-time PCR screening test for the strain-specific detection of campylobacters in environmental samples has been developed to address this issue. To enable this approach, fluorescently labeled PCR oligonucleotide probes suitable for a LightCycler-based assay were designed to match a highly variable DNA segment within the flaA short variable region (SVR) of Campylobacter jejuni or C. coli. The capacity of such probes to provide strain-specific tools was investigated by using bacterial cultures and spiked and naturally contaminated poultry fecal and environmental samples. The sensitivity of two representative probes was estimated, by using two different C. jejuni strains, to be 1.3 x 10(2) to 3.7 x 10(2) CFU/ml in bacterial cultures and 6.6 x 10(2) CFU/ml in spiked fecal samples. The specificity of the SVR for C. jejuni and C. coli was confirmed by using a panel of strains comprising other Campylobacter species and naturally contaminated samples. The approach was field tested by sampling the environment and feces of chickens of two adjacently located poultry houses on a conventional broiler farm throughout the life of one flock. All environmental samples were enriched for 2 days, and then DNA was prepared and stored. Where feasible, campylobacter isolates were also recovered and stored for subsequent testing. A strain-specific probe based on the SVR of the strain isolated from the first positive chicken fecal sample was developed. This probe was then used to screen the stored environmental samples by real-time PCR. Pulsed-field gel electrophoresis was used to compare recovered environmental and fecal isolates to assess the specificity of the method. The results established the proof of principle that strain-specific probes, based on the SVR of flaA, can identify a flock-colonizing strain in DNA preparations from enriched environmental cultures. Such a novel strategy provides the opportunity to investigate the epidemiology of campylobacters in poultry flocks and allows targeted biosecurity interventions to be developed. The strategy may also have wider applications for the tracking of specific campylobacter strains in heavily contaminated environments.

Project description:In order to survive, animals must quickly and accurately locate prey, predators, and conspecifics using the signals they generate. The signal source location can be estimated using multiple detectors and the inverse relationship between the received signal intensity (RSI) and the distance, but difficulty of the source localization increases if there is an additional dependence on the orientation of a signal source. In such cases, the signal source could be approximated as an ideal dipole for simplification. Based on a theoretical model, the RSI can be directly predicted from a known dipole location; but estimating a dipole location from RSIs has no direct analytical solution. Here, we propose an efficient solution to the dipole localization problem by using a lookup table (LUT) to store RSIs predicted by our theoretically derived dipole model at many possible dipole positions and orientations. For a given set of RSIs measured at multiple detectors, our algorithm found a dipole location having the closest matching normalized RSIs from the LUT, and further refined the location at higher resolution. Studying the natural behavior of weakly electric fish (WEF) requires efficiently computing their location and the temporal pattern of their electric signals over extended periods. Our dipole localization method was successfully applied to track single or multiple freely swimming WEF in shallow water in real-time, as each fish could be closely approximated by an ideal current dipole in two dimensions. Our optimized search algorithm found the animal's positions, orientations, and tail-bending angles quickly and accurately under various conditions, without the need for calibrating individual-specific parameters. Our dipole localization method is directly applicable to studying the role of active sensing during spatial navigation, or social interactions between multiple WEF. Furthermore, our method could be extended to other application areas involving dipole source localization.

Dataset Information

A modular approach to integrating multiple data sources into real-time clinical prediction for pediatric diarrhea.

Publications

A modular approach to integrating multiple data sources into real-time clinical prediction for pediatric diarrhea.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets