Clarifying the effect of library batch on extracellular RNA sequencing

We read with great interest the recent paper of Zhou et al. (1) which describes a promising low-input protocol for measuring secreted RNA in blood. Zhou et al. (1) apply this technology to 96 samples of serum from cancer patients (28 with recurrence, 68 without) and 32 samples of serum from healthy controls. In the face of high noise (r2 ∼ 0.5 over 12 log-orders), the cohorts can be distinguished with an area under the receiver operating characteristic curve of >0.95 on the basis of extracellular RNA. This classification performance is stronger than what has been observed in circulating tumor cells (2), so we sought to understand the nature of the extracellular cancer signal. … [↵][1]1To whom correspondence may be addressed. Email: gary.gao{at}singleragenomics.com. [1]: #xref-corresp-1-1