A major challenge with respect to comprehensive interpretation is the variable composition of different cell types in a given sample of inflamed tissue or peripheral blood. All discovery driven studies referenced above were performed with such mixed samples. It is a problem of almost every clinical sample in any type of disease, including tumour tissues [17-20].
We have compared the impact of disease on one specific cell type with the differences between a normal cell type and a normal tissue. Figure 1 demonstrates that there are many more genes and to a higher extent differentially expressed, for example between normal monocytes and normal synovium, than between normal mono-cytes and rheumatoid arthritis monocytes.
Variability in cellular composition will therefore have a major impact on the quality and quantity of differentially expressed genes. This will increase statistical variability, will be more prone to false selection in cases of a small number of samples and will therefore demand for larger sample sizes as suggested for the cancer studies . Another consequence of mixed cellular composition is that genes related to infiltration of a cell type may be preferably selected, while genes involved in the molecular processes of the disease may be lost in the majority of minor differences. As a third aspect, we have to consider that many genes provide their function not by upregulation of transcription but by protein modification. Thus, the message will be detectable in a normal and un-stimulated cell. If such a gene is known as a potent signal transmitter, infiltration of the corresponding cell type will pretend active participation of the gene in the disease process, although, this may not be true. Finally, mixed cell types dilute the message of a single cell type and rare transcripts may fall below the threshold of detection.
These problems are inherent to all studies performed with samples of complex cellular composition. In the different studies referenced above, only one single group  tried to address this problem and identified differential cellular composition as the source of a certain panel of differentially expressed genes. When we compared the list of candidates published, for example in another study on PBMC , with our own expression profiles of different highly purified subtypes of peripheral blood cells, all genes were found to be expressed highest in neutrophils. In light of the first study  that such cells may be co-separated in inflammatory diseases, the data published by the second study  have to be considered with caution and need further confirmation.
Was this article helpful?