Estrogen receptor status prediction by gene component regression: a comparative study.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Microarray studies generate high-dimensional gene expression data.
• High dimensionality and collinearity pose challenges in analyzing microarray data.
• Accurate tumor classification is crucial for effective breast cancer treatment.

Purpose of the Study:

• To evaluate gene component analysis for reducing dimensionality in microarray data.
• To assess the effectiveness of different dimensionality reduction strategies (PCR, PLS, RRR) for tumor classification.
• To investigate the impact of gene selection on classification performance.

Main Methods:

• Applied Principle Component Regression (PCR), Partial Least Square (PLS), and Reduced Rank Regression (RRR) to breast cancer microarray data.
• Utilized Logistic Regression (LR) and Linear Discriminative Analysis (LDA) for tumor classification using derived gene components.
• Evaluated the influence of gene selection and filtration techniques.

Main Results:

• Gene component analysis successfully reduced data dimensionality and addressed collinearity issues.
• Classifiers based on gene components accurately discriminated Estrogen Receptor (ER) positive from ER negative breast cancers.
• The developed classifiers demonstrated high predictive accuracy when validated on an independent microarray dataset.

Conclusions:

• Gene component analysis is a viable strategy for managing high-dimensional gene expression data in microarray studies.
• Dimensionality reduction techniques can enhance the performance of tumor classification models.
• The proposed gene component classifiers offer a robust and accurate approach for breast cancer subtyping.