RCRnorm: An integrated system of random-coefficient hierarchical regression models for normalizing NanoString nCounter data

Abstract

Formalin-fixed paraffin-embedded (FFPE) samples have great potential for biomarker discovery, retrospective studies and diagnosis or prognosis of diseases. Their application, however, is hindered by the unsatisfactory performance of traditional gene expression profiling techniques on damaged RNAs. NanoString nCounter platform is well suited for profiling of FFPE samples and measures gene expression with high sensitivity which may greatly facilitate realization of scientific and clinical values of FFPE samples. However, methodological development for normalization, a critical step when analyzing this type of data, is far behind. Existing methods designed for the platform use information from different types of internal controls separately and rely on an overly-simplified assumption that expression of housekeeping genes is constant across samples for global scaling. Thus, these methods are not optimized for the nCounter system, not mentioning that they were not developed for FFPE samples. We construct an integrated system of random-coefficient hierarchical regression models to capture main patterns and characteristics observed from NanoString data of FFPE samples and develop a Bayesian approach to estimate parameters and normalize gene expression across samples. Our method, labeled RCRnorm, incorporates information from all aspects of the experimental design and simultaneously removes biases from various sources. It eliminates the unrealistic assumption on housekeeping genes and offers great interpretability. Furthermore, it is applicable to freshly frozen or like samples that can be generally viewed as a reduced case of FFPE samples. Simulation and applications showed the superior performance of RCRnorm.

An approach for normalization and quality control for NanoString RNA expression data

“…We benchmarked RUVSeq and nSolver with two other normalization methods, NanoStringDiff [13] and 2 RCRnorm [14]. We observed differences across the four normalization strategies (described in 2 Supplemental Table S2), namely greater remaining technical variation using nSolver and NanoStringDiff 2 than RCRnorm and RUVSeq (Figure 2B-D).…”

RCRdiff: A fully integrated Bayesian method for differential expression analysis using raw NanoString nCounter data

“…However, information provided by different types of probes is intermingled. Instead of correcting the different biases in an isolated manner, which is typical in existing methods, Jia et al 4 proposed an integrated system of random‐coefficient regression models for jointly modeling log‐transformed read counts from the different types of probes. For the i th sample (), let , , , and denote the transformed read count of the p th positive control (), the n th negative control (), the h th housekeeping gene (), and the r th regular gene , respectively.…”

“…Using information from the different types of internal controls, several analysis methods have been developed for the nCounter platform to normalize and extract gene expression levels, such as NanoStringNorm, 1 NAPPA, 2 and NanoStringDiff 3 . Recently, Jia et al 4 proposed a novel Bayesian method called RCRnorm, which consists of a system of random‐coefficient hierarchical regression (RCR) models, attempting to accurately characterize different types of probes. It has been shown that RCRnorm compares favorably with other existing methods, especially for situations with an elevated level of heterogeneity from various sources.…”

“…We aim to integrate Bayesian LASSO into the system of RCR models 4 to identify the differentially expressed (DE) genes from NanoString nCounter data. Bayesian LASSO methods can provide not only closer parameter estimates as compared to original LASSO 8 but also reliable estimates for standard errors that can measure model uncertainty 9,10 .…”

An approach for normalization and quality control for NanoString RNA expression data

“…Using CBCS data, we compared the normalized datasets from nSolver, RUVSeq, NanoStringDiff [13], 1 and RCRnorm [14] with the raw data through visualization methods outlined above (Figure 1 Table S2. 1 1 that RUVSeq best addressed the largest source of technical variation identified in the raw data ( Figure 3 2D; Supplemental Figure S5A) while also not removing a significant portion of biological variation 3 (Supplemental Figure S5B).…”