A robust RUV-testing procedure via gamma-divergence

TL;DR

This paper introduces a robust RUV-testing method using gamma-divergence to effectively remove unwanted variation and outliers in gene expression data, improving the detection of differentially expressed genes.

Contribution

The proposed method is model-free for outliers, easy to implement with a tuning parameter, and includes a data-driven approach for parameter selection.

Findings

Successfully removes unwanted variation in gender study data

Identifies more DE-genes than conventional methods

Demonstrates robustness against outliers

Abstract

Identification of differentially expressed genes (DE-genes) is commonly conducted in modern biomedical researches. However, unwanted variation inevitably arises during the data collection process, which could make the detection results heavily biased. It is suggested to remove the unwanted variation while keeping the biological variation to ensure a reliable analysis result. Removing Unwanted Variation (RUV) is recently proposed for this purpose by the virtue of negative control genes. On the other hand, outliers are frequently appear in modern high-throughput genetic data that can heavily affect the performances of RUV and its downstream analysis. In this work, we propose a robust RUV-testing procedure via gamma-divergence. The advantages of our method are twofold: (1) it does not involve any modeling for the outlier distribution, which is applicable to various situations, (2) it is easy to implement in the sense that its robustness is controlled by a single tuning parameter gamma of gamma-divergence, and a data-driven criterion is developed to select $\gamma$. In the Gender Study, our method can successfully remove unwanted variation, and is able to identify more DE-genes than conventional methods.