The Value of Multi-Omics Combination in the Diagnosis of Specific Sepsis

Objective A machine learning classification algorithm was employed to construct a high-performance early diagnosis model for sepsis infection with specific pathogenic bacteria based on multi-omics data. Then we compared the prediction effect between the single-omics model and the multi-omics model. Methods This was a secondary analysis of two observational studies. The omics data was extracted and integrated. Support vector machine （SVM） algorithm was used to construct three prediction models whose performance were compared mutually. Results The multi-omics model showed the best performance （Staphylococcus aureus bacteria （SaB） vs. others, （area under the receiver operating characteristic curve, AUC）=0.97; non_SaB vs. others, AUC=0.94; Control vs. others, AUC=0.94 comparing with single-omic model. Conclusions Multi-omics prediction model had tremendous potential in identifying specific sepsis and performed better than single-omic model.