Random forest classification of four unsaturated fatty acids for infection diagnosis and prognosis in female Chinese patients.

Objective: Unsaturated fatty acids (UFAs) are important in immune regulation and inflammation; however, their diagnostic and prognostic value in infected patients is unclear. This study evaluated UFAs in infected female patients versus healthy controls using random forest (RF) analysis. Methods: A total of 115 female subjects, consisting of healthy controls and infected patients, were recruited. Clinical indices and UFAs, arachidonic acid (AA), docosahexaenoic acid (DHA), linoleic acid, and alpha-linolenic acid (ALA), and complete blood cell (CBC) count data were analyzed. Correlation and regression analyses were conducted to assess the relationships between UFAs in healthy and infected states. Receiver operating characteristic (ROC) analysis and orthogonal partial least squares discriminant analysis (OPLS-DA) modeling were performed to evaluate the diagnostic value of UFAs. An RF model was developed to classify infected and healthy populations. Results: Significant differences were observed in CBC and UFA indices between healthy and infected female patients. The ROC analysis demonstrated that UFAs showed statistically significant differences, although these indices alone did not completely discriminate between infected and healthy individuals. However, the RF model, incorporating both UFAs and CBC data, achieved a clear separation between the infected and healthy groups, outperforming the OPLS-DA model. The ROC results for both UFAs and CBC datasets indicated predictive value for infection prognosis. Conclusion: The integration of UFAs and CBC data within an RF model provides enhanced diagnostic and prognostic accuracy compared to OPLS-DA. UFAs, particularly AA and ALA, possess substantial predictive value for infection prognosis.