Pb2573: microrna clusters facilitate the prediction of infectious fever after hematopoietic stem cell transplantation

Topic: 30. Infections in hematology (incl. supportive care/therapy) Background: Infectious fever is a common problem with high morbidity post hematopoietic stem cell transplantation (HSCT). Aims: To identify microRNAs related to infectious diseases and clusters with values of prediction or diagnosis for infectious fever using machine learning algorithm. Methods: Totally, 345 serum and 56 bronchoalveolar lavage fluid (BALF) samples were collected and the infectious fever classifiers based on microRNA clusters are established. Results: MicroRNA clusters, including miR-155-5p, miR-4496, miR-183-5p, miR-99a-5p, miR-769-5p, miR-23a-5p, miR-22-5p and miR-9-5p, were calculated out to be the diagnosis markers by using the Random Forest (RF) and Support Vector Machine (SVM) algorithms, and the AUCs of the validating serum samples were 0.97 and 0.99 respectively. The highest prediction accuracy of the linear regression, random forest and SVM algorithms acquired a 20-30 microRNAs cluster. Through the time-pointed linear regression analysis of infectious fever, a cluster constituted with 31 microRNAs has the highest fit effect (R2=0.81) and distinguishes the infectious and noninfectious fever in untrained serum samples at 0-2 weeks ahead with an accuracy of 87%. Moreover, the correlation analysis of serum microRNAs and hematopoietic reconstitution or immune response identified the specific clusters corresponding to infections after HSCT. Summary/Conclusion: Overall, the combination of serum microRNAs and machine learning algorithms displayed a promising potential in the prediction and diagnosis of infectious fever after HSCT. Keywords: HSCT, Fever, Immune response, Infection