Missing value imputation using shrinkage regression method in microarray data

Microarray experiment technique is an important and powerful tool for biomedical research. Although the microarray technology has been developed for more than a decade, more than 90% of genes in microarray data are contained missing value. Inaccurate missing value estimation reduces the power of the microarray data anal-ysis algorithms. Many types of methods are used to estimate missing values. Among them, the regression-based methods as Local Least Square impute (LLSimpute) have been performed better than the other types of methods. Shrinkage LLSimpute method proposed to further improve the performances of the LLSimpute. Shrinkage LL-Simpute take advantage of the correlation structure of the microarray data and choose similar genes for the target gene (with missing values) by Pearson correlation coefficients. In order to adjust the coefficients of the regres-this method combined the least squares principle with shrinkage estimation, and then used the new ,sion model coefficients to estimate missing values. Results show that Shrinkage LLSimpute provided more accurate missing value estimation than LLSimpute in microarray dataset (zinc homeostasis, Lyons et al., 2000). Missing value imputation is as a necessary pre-processing step in the microarray data analysis, because, the most of analysis algorithms, such as clustering and network require a complete dataset. Hence, accurate and efficient methods for imputation missing values have become an essential issue. Since shrinkage LLSimpute method can provide accurate missing value estimation, it is competitive alternative to the LLSimpute method

کلیدواژه ها:

Imputation ، Missing Values ، Microarray ، Regression ، Shrinkage Estimate