Generalized beta convolution model of the true intensity for the Illumina BeadArrays

TL;DR

This paper introduces a generalized beta convolution model for background correction in Illumina BeadArrays, extending existing models with more flexible distribution assumptions to improve noise reduction in microarray data.

Contribution

It proposes a new generalized beta convolution model for background correction, broadening the class of models used in microarray data preprocessing.

Findings

Enhanced background correction accuracy demonstrated

Flexible modeling of noise distributions achieved

Improved microarray data quality inferred

Abstract

Microarray data come from many steps of production and have been known to contain noise. The pre-processing is implemented to reduce the noise, where the background is corrected. Prior to further analysis, many Illumina BeadArrays users had applied the convolution model, a model which had been adapted from when it was first developed on the Affymetrix platform, to adjust the intensity value: corrected background intensity value. Several models based on different underlying distributions and or parameters estimation methods have been proposed and applied. For instance : the exponential-gamma, the normal-gamma and the exponential-normal convolutions with a maximum likelihood estimation, non-parametric, Bayesian and moment methods of the parameters estimation, including two recent exponential-lognormal and gamma-lognormal convolutions. In this paper, we propose models and derive the corrected background intensity based on the generalized betas and the generalized beta-normal convolutions as a generalization of the existing models.