Variable selection methods for developing a biomarker panel for prediction of dengue hemorrhagic fever

Abstract

BackgroundThe choice of selection methods to identify important variables for binary classification modeling is critical to produce stable models that are interpretable, that generate accurate predictions and have minimum bias. This work is motivated by data on clinical and laboratory features of severe dengue infections (dengue hemorrhagic fever, DHF) obtained from 51 individuals enrolled in a prospective observational study of acute human dengue infections.ResultsWe carry out a comprehensive performance comparison using several classification models for DHF over the dengue data set. We compared variable selection results by Multivariate Adaptive Regression Splines, Learning Ensemble, Random Forest, Bayesian Moving Averaging, Stochastic Search Variable Selection, and Generalized Regularized Logistics Regression. Model averaging methods (bagging, boosting and ensemble learners) have higher accuracy, but the generalized regularized regression model has the highest predictive power because the linearity assumptions of candidate predictors are strongly satisfied via deviance chi-square testing procedures. Bootstrapping applications for evaluating predictive regression coefficients in regularized regression model are performed.ConclusionsFeature reduction methods introduce inherent biases and therefore are data-type dependent. We propose that these limitations can be overcome using an exhaustive approach for searching feature space. Using this approach, our results suggest that IL-10, platelet and lymphocyte counts are the major features for predicting dengue DHF on the basis of blood chemistries and cytokine measurements.

Dengue NS1 antigen contributes to disease severity by inducing interleukin (IL)-10 by monocytes

“…Our previous studies and studies conducted by others have shown that serum IL-10 was associated with severe disease [7,39,40]. In addition, we reported previously that IL-10 inhibits DENV-specific T cell responses [8].…”

Improved Detection of Invasive Pulmonary Aspergillosis Arising during Leukemia Treatment Using a Panel of Host Response Proteins and Fungal Antigens

“…To identify the best method for combining the differentially expressed proteins, we evaluated a group of machine learning classifiers for performance in the IPA discovery/qualification data set using 10-fold cross validation. Recognizing that the data were not parametrically distributed, and the performance of machine learning classifiers are highly dependent on the underlying data structures [ 22 , 25 ], we evaluated the performance of classification and regression trees (CART), random forests (RF), multivariate regression spine (MARS), and generalized pathseeker (GPS) machine learning techniques by their classification accuracy and area under the ROC curve (AUC). Although all classifiers evaluated generated accurate models on the training data set, there were significant differences in their performance on the verification data set ( Table 6 ).…”

Fc receptors in antibody‐dependent enhancement of viral infections

“…Upregulated TNF, IL‐6, and IL‐10 have been reported in several human investigations on DENV infection . Intriguingly, IL‐10 has recently been considered to be a reliable biomarker for predicting clinical severity . Although data for FcγR‐mediated cytokine responses in the context of DENV ADE are variable , increased TNF and IL‐6 production has consistently been shown in monocytes, macrophages, and DCs upon exposure to immunocomplexes .…”

“…Thus, it is not clear whether the inhibition of Syk by LILRB‐1 may lead to a reduction in IL‐10 production under ADE conditions, but this may involve a temporal balance of inhibition and activation of Syk yet to be elucidated. If so, based on data from patients with severe DENV infection , this balance is presumably tuned in favor of excessive IL‐10 production.…”