Abstract
RNA ISH (Ribonucleic acid in situ hybridization) assays are an ever-expanding application due to the ability to evaluate molecular targets, while retaining tissue morphology. The rate limiting step in RNA ISH assays is the time consuming and error prone method of manually counting signal under a microscope. The Aperio RNA ISH Algorithm offers a reproducible, fast, and quantitative method of evaluating tissue samples that have been stained to detect RNA ISH signal. This single algorithm can be used on numerous tissue types for both single and dual-plex assays.

In this paper, we describe a validation study that was performed to verify the correlation between the Aperio RNA ISH Algorithm and the current gold standard method of manual interpretation. A total of 30 digital slides ranging in tissue source and assays types were scored manually by a scientist and the resulting data were correlated with scores obtained from the Aperio RNA ISH Algorithm. In both modalities, the number of cells, count of signal within the cell, and signal in all tissue were recorded. The high level of correlation between the two methods (R2 >0.99) confirms that automated image analysis can be used as a fast and reproducible alternative to the traditional methods of manual interpretation.
Introduction
RNA ISH is a rapidly growing method for the analysis of molecular targets within tissue samples. It enables identification of individual copies of targets, while maintaining tissue morphology, a feature that is lost in other methods such as PCR1. RNA ISH technology is being used in many areas of cancer research today2-5. In addition, there is the potential benefit of being able to combine RNA ISH assays with traditional IHC assays, thereby enabling users to visualize both RNA and protein status on a single slide6. Manual interpretation of RNA ISH signal is time-consuming and typically uses the semi-quantitative approach of ordinal scores (0, 1+, 2+, 3+), which often mask discrete cohorts that are readily identifiable when fully quantitative analysis is performed. Moreover, manual reads are subject to inter- and intra-observer variability, resulting in a lack of reproducibility and standardization in staining interpretation7–10.