Abstract
Background
DNA microarrays provide informative data for transcriptional profiling and identifying gene expression signatures to help prevent progression from latent tuberculosis infection (LTBI) to active disease. However, constructing a prognostic model for distinguishing LTBI from active tuberculosis is very challenging due to the noisy nature of data and lack of a generally stable analysis approach.
Methods
In the present study, we proposed an accurate predictive model with the help of data fusion at the decision level. In this regard, results of filter feature selection and wrapping feature selection techniques were combined with multiple-criteria decision-making (MCDM) methods to select 26 genes from six microarray datasets that can be the most distinctive genes for diagnosing tuberculosis cases. As the main contribution of this study, the final ranking function was constructed by combining protein-protein interaction (PPI) network with an MCDM method (DEMATEL) to improve our feature ranking approach pointedly.
Results
The best fusion of classifiers is determined to detect different types of tuberculosis with a sensitivity of 0.949514, specificity of 0.885872 and accuracy of 0.922368. By applying data fusion at the decision level on the 26 introduced genes in terms of fusion of classifiers of random forests (RF) and k-nearest neighbors (KNN) regarding Yager's theory, the proposed algorithm reached an accuracy of 0.922368. Finally, with the help of cumulative clustering, the pairs of genes involved in the diagnosis of latent and activated tuberculosis have been introduced.
Conclusions
So far, a combination of MCDM methods and PPI networks have not been used to diagnose different states of tuberculosis.