Collaborative representation-based classification of microarray gene expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray technology enables simultaneous gene expression analysis over time.
• Sparse representation (SR)-based methods are used for classifying gene expression data by clustering.
• Existing methods face challenges in complexity and classification accuracy.

Purpose of the Study:

• To propose a novel collaborative representation (CR)-based classification method for gene expression data.
• To enhance classification accuracy and reduce computational complexity.
• To validate the method's efficacy in disease subtype detection.

Main Methods:

• Developed a CR-based classification approach using regularized least squares.
• Employed compressive sensing for dimensionality reduction of high-dimensional gene expression data.
• Classified testing samples by minimizing representation error based on training samples.

Main Results:

• The CR-based method demonstrated significantly higher stability and accuracy than traditional classifiers like Support Vector Machine (SVM).
• Compressive sensing achieved lossless compression, reducing computational load without information loss.
• Experiments on leukemia and autism spectrum disorder subtypes showed promising results.

Conclusions:

• The proposed CR-based algorithm is a robust and accurate method for gene expression data classification.
• The integration of compressive sensing effectively addresses the challenge of high-dimensional data.
• This approach holds potential for improved disease subtyping and biomarker discovery.