| Summary: | Sepsis is a potentially life-threatening medical condition that increases mortality in pediatric populations admitted in the intensive care unit (ICU). Due to the unpredictable nature of the disease course, it was challenging to find the informative genetic biomarkers at the earliest stages. Consequently, a considerable attention has been paid for the early prediction of pediatric sepsis based on genetic biomarkers analysis that would promote the early medical intervention. Therefore, the proposed study attempted to demonstrate the feasibility of Henry Gas Solubility Optimization (HGSO) in differential gene selection to train supervised machine learning algorithms for the early prediction of pediatric sepsis and survival rate evaluation. 26 nonoverlapping informative genes have been nominated using the gene expression profile of peripheral blood cells. After 20 runs of 5-fold cross-validation, the selected genes revealed its effectiveness in the early identification of sepsis subtypes with an estimated average accuracy of 98.03 ± 0.30 % evaluated using 20 runs of fivefold cross-validation and an average accuracy of 98.83 ± 0.57 % for evaluating the survival rate. Based on the experimental results, the present study using the novel metaheuristic algorithm HGSO determined the highest accuracy, the most predictive and informative genes for pediatric sepsis, thus allowing determination of the appropriate treatment plan.
|
|---|