Smoothing blemished gene expression microarray data via missing value imputation.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Gene expression microarray data is crucial for biological and medical research.
• Microarray data is susceptible to noise from experimental factors, impacting reliability.
• Accurate data is essential for robust biomarker identification and disease diagnosis.

Purpose of the Study:

• To develop a computational method for identifying and correcting inaccurate data entries ('stains') in gene expression microarrays.
• To improve the quality of microarray data for downstream analyses.
• To enhance the accuracy of biomarker discovery and disease diagnosis using smoothed data.

Main Methods:

• A novel computational approach combining feature gene selection and sample classification.
• Identification of a small percentage (1-4%) of likely inaccurate data entries ('stains').
• Application of imputation methods to estimate and correct the identified 'stains'.

Main Results:

• The proposed method effectively identifies and corrects 'stains' in microarray data.
• Smoothed datasets derived from the method show improved quality for analysis.
• Experimental results on Affymetrix platforms demonstrated enhanced biomarker identification and cancer diagnosis.

Conclusions:

• The developed computational method offers an efficient way to curate noisy gene expression microarray data.
• Data smoothing using this method leads to more reliable downstream analyses.
• This approach has significant implications for improving accuracy in biological and medical research, particularly in cancer diagnosis.