An integrated approach for identifying wrongly labelled samples when performing classification in microarray data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Hybrid approaches for gene selection and classification offer improved results over independent methods.
• Existing methods struggle with microarray data accuracy and gene set stability due to potential sample outliers.
• Current outlier detection algorithms are often unsuitable for microarray data or address only outlier detection.

Purpose of the Study:

• To address limitations in hybrid microarray analysis by incorporating robust outlier detection.
• To enhance classification accuracy and gene set stability in microarray datasets.
• To develop an improved Multiple-Filter-Multiple-Wrapper (MFMW) model for outlier detection.

Main Methods:

• Proposed an enhanced MFMW-outlier model integrating outlier detection into a hybrid framework.
• Implemented an unbiased external Leave-One-Out Cross-Validation for improved model evaluation.
• Utilized an L1-norm Support Vector Machine (SVM) for stable gene selection after outlier removal.

Main Results:

• The MFMW-outlier model successfully identified all known outliers in six benchmark microarray datasets.
• Selected genes demonstrated biological relevance, and classification accuracy was improved post-outlier removal.
• Outperformed PRAPIV on synthetic datasets, achieving better average precision and recall.

Conclusions:

• The MFMW-outlier approach effectively detects outliers in high-dimensional microarray data.
• This method enhances the reliability of gene selection and classification in biological studies.
• The proposed model offers a powerful tool for improving the quality of microarray data analysis.