Missing value imputation for microarray gene expression data using histone acetylation information.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Accurate estimation of missing values in microarray data is essential for downstream bioinformatics analyses.
• Existing imputation methods often perform poorly on datasets with high percentages of missing values.

Purpose of the Study:

• To explore the feasibility of using gene regulatory mechanisms for missing value imputation in microarray data.
• To develop and evaluate a novel imputation framework incorporating histone acetylation information.

Main Methods:

• Developed the histone acetylation information aided imputation method (HAIimpute).
• Integrated histone acetylation data into K-nearest neighbor (KNN) and local least square (LLS) imputation algorithms.
• Evaluated performance using normalized root mean squared error (NRMSE) and Pearson correlation coefficients.

Main Results:

• HAIimpute methods demonstrated significant improvements in microarray imputation accuracy compared to conventional KNN and LLS.
• Imputed genes using HAIimpute showed higher correlation with original complete genes.
• The proposed methods outperformed GOimpute, a related method using functional similarity.

Conclusions:

• Histone acetylation information substantially enhances imputation performance, particularly for high missing data percentages.
• The proposed approach can be generalized to other imputation methods.
• Further improvements are expected with the integration of additional gene regulatory mechanism knowledge.