Methodology to identify a gene expression signature by merging microarray datasets.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Microarray datasets are crucial for identifying differentially expressed genes and gene expression signatures.
• Limited sample sizes in individual datasets reduce statistical power and generalizability.
• Merging datasets is challenging but necessary to overcome sample size limitations.

Purpose of the Study:

• To develop and validate a methodology for merging microarray datasets to identify reliable gene expression signatures.
• To improve the statistical, predictive, and generalization power of gene expression analysis.
• To aid in disease diagnosis, prognosis, and prediction of therapeutic response.

Main Methods:

• A novel methodology combining statistical methods and supervised machine learning algorithms for merging microarray datasets.
• Statistical methods were used to identify differentially expressed genes across datasets.
• Supervised machine learning was employed to select the final gene expression signature.

Main Results:

• The methodology was successfully validated on heart failure and autism spectrum disorder microarray datasets.
• For heart failure, a 117-gene signature achieved ~98% classification accuracy.
• For autism spectrum disorder, a 79-gene signature achieved ~82% classification accuracy.

Conclusions:

• The proposed methodology effectively merges microarray datasets to identify robust gene expression signatures.
• This approach enhances the accuracy of disease classification and biomarker discovery.
• The R-implemented methodology is publicly available, facilitating further research.