Fig. 4: Estimate biological features of a new dataset using the signatures learned from public databases.

RAVs encompass biological signals applicable across different platforms and independent datasets. We demonstrate this transfer learning capacity of RAVs by identifying the neutrophil-associated RAV from systemic lupus erythematosus whole blood (SLE-WB) data and using the same RAV to analyze nasal brushing (NARES) dataset. a Neutrophil counts of 853 samples from the SLE-WB dataset were plotted against RAV1551-assigned sample scores. b Neutrophil count estimates by MCPcounter were plotted against sample scores assigned by RAV1551. c Neutrophil count of 76 NARES samples were estimated by MCPcounter and plotted against RAV1551-assigned sample scores. The shaded area is the 95% confidence interval for predictions from a linear model.