Biological impact of missing-value imputation on downstream analyses of gene expression profiles

Abstract

Motivation:Microarray experiments frequently produce multiple missing values (MVs) due to flaws such as dust, scratches, insufficient resolution or hybridization errors on the chips. Unfortunately, many downstream algorithms require a complete data matrix. The motivation of this work is to determine the impact of MV imputation on downstream analysis, and whether ranking of imputation methods by imputation accuracy correlates well with the biological impact of the imputation. Methods: Using eight datasets for differential expression (DE) and classification analysis and eight datasets for gene clustering, we demonstrate the biological impact of missing-value imputation on statistical downstream analyses, including three commonly employed DE methods, four classifiers and three gene-clustering methods. Correlation between the rankings of imputation methods based on three root-mean squared error (RMSE) measures and the rankings based on the downstream analysis methods was used to investigate which RMSE measure was most consistent with the biological impact measures, and which downstream analysis methods were the most sensitive to the choice of imputation procedure. Results: DE was the most sensitive to the choice of imputation procedure, while classification was the least sensitive and clustering was intermediate between the two. The logged RMSE (LRMSE) measure had the highest correlation with the imputation rankings based on the DE results, indicating that the LRMSE is the best representative surrogate among the three RMSE-based measures. Bayesian principal component analysis and least squares adaptive appeared to be the best performing methods in the empirical downstream evaluation.

IMDE: An easy-to-use web server for missing data estimation

“…Previous studies have demonstrated that the performance of an imputation algorithm could be affected by many factors, hence no algorithm can perform well on all kinds of datasets [5], [6]. This phenomenon can also be observed using IMDE.…”

“…Second, NRMSE is the only performance index used in IMDE. Other biologically relevant performance indices such as the Conserved Pair Proportion (CPP) [3] and Biomarker List Concordance Index (BLCI) [5] should also be used. We will incorporate them in the future version of IMDE.…”

“…Moreover, some studies [2], [3], [4] had indicated that results of MV imputation would strongly impact the accuracy of microarray data analysis. Other studies [5], [6], [7] had reported that the performance of an imputation algorithm depends on the types of datasets, suggesting that MVs should be cautiously treated. Therefore, MV imputation has become an indispensable step during the preprocessing stage of the analysis of high-throughput data.…”

Bayesian model-based clustering of temporal gene expression using autoregressive panel data approach

“…For example, the values assumed for φ 1 k in the simulation were the average values of the estimates obtained from each resulting cluster x i ∈ S j Y i −S j 2 , where S j is the mean of point in S j . The choice of k-means is due to its general use in gene expression clustering (Mar et al, 2011;Oh et al, 2011), but it is important emphasize that the success of this method depends of the stated number of clusters, which was the same number used in the simulation study (five), i.e. was considered the optimum condition for the application of this method.…”

A systems biology approach identifies a gene regulatory network in parotid acinar cell differentiation.

“…expression was not detected in seven or more samples, which excluded 100 microRNAs), missing values from remaining probes were imputed using R function imputeKNN in Bioconductor package MmPalateMicroRNA [119] (14.5% of values were imputed) [120]. A median sweep was performed to normalize delta Ct values by subtracting the global median for each array.…”