Statistical Inferences of HIVRNA and Fracture Based on the PAK1 Expression via Neural Network Model

Abstract

Background: Acquired immune deficiency syndrome and fracture are all serious hazards to human health that create a widespread alarm. Biomarkers that are closely linked to HIVRNA and fracture are unknown. Methods: 48 cases with HIV and fracture and 112 normal cases were recruited. Blood neutrophil count (NEU), white blood cell count (WBC), PAK1 and HIVRNA were measured. Pearson's chisquared test was used to evaluate the association between HIVRNA with fracture and NEU, WBC, PAK1. BP neural network model was constructed to analyze the predictive power of the combined effects of NEU, WBC, PAK1 for HIV RNA with fracture. Results: There exist strong correlations between PAK1, NEU, WBC and HIVRNA with fracture. The neural network model was successfully constructed. The overall determination coefficients of the training sample, validation sample, and test sample were 0.7235, 0.4795, 0.6188, 0.6792, respectively, indicating that the fitting effect between training sample and overall was good. Statistical determination coefficient of the goodness of fit R2 ≈ 0.82, it can be considered that degree of fit between the estimate and corresponding actual data is good. Conclusion: HIVRNA with fracture could be predicted using a neural network model based on NEU, WBC, PAK1. The neural network model is an innovative algorithm for forecasting HIVRNA levels with fracture.