Sepsis is a common and serious health problem whereby improvements in diagnosis are crucial in increasing survival rates. To test whether profiling transcription is applicable to sepsis diagnosis, we analyzed whole blood using a microarray containing probes for 340 genes relevant to inflammation. The patient's gene expression pattern was highly homogenous, resulting in 69% of differentially expressed genes. With a positive predictive value of 98%, a list of 50 differentially expressed genes was compiled, and randomly chosen transcripts were confirmed by PCR. Here, we present the first evidence that microarrays can identify typical gene expression profiles in the blood of patients with severe sepsis. Regardless of the heterogeneity of the patients, we observed a striking correlation between the conventional diagnostic classification and our approach. The unity of responses suggests that the principle of this multiparameter approach can be adapted to early stage sepsis diagnosis.