Related Experiment Video
Updated: Aug 9, 2026
Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps
Published on: October 28, 2018
Molecular classification of cancer: unsupervised self-organizing map analysis of gene expression microarray data
David G Covell1, Anders Wallqvist, Alfred A Rabow
1National Cancer Institute-Frederick, Science Applications International Corporation-Frederick, Developmental Therapeutics Program, Screening Technologies Branch, Laboratory of Computational Technologies, Frederick, Maryland 21702, USA.
This study introduces a self-organizing map (SOM) method for classifying cancer types using gene expression data. The SOM approach effectively distinguishes tumor from normal tissues and achieves high accuracy for several cancer types.
Area of Science:
- Bioinformatics
- Computational Biology
- Genomics
Background:
- Gene expression profiling using oligonucleotide microarrays is crucial for cancer research.
- Accurate classification of diverse tumor types remains a challenge in bioinformatics.
- Developing robust computational methods is essential for analyzing complex gene expression datasets.
Purpose of the Study:
- To develop and evaluate an unsupervised self-organizing map (SOM) based clustering strategy for classifying tissue samples.
- To identify gene expression patterns that can distinguish between different tumor classes and normal tissues.
- To assess the accuracy and limitations of the SOM method in tumor classification.
Main Methods:
- Utilized an unsupervised self-organizing map (SOM) algorithm for clustering gene expression data from an oligonucleotide microarray patient database.
- Applied the SOM to classify tissue samples into distinct tumor types and normal controls.
- Analyzed classification accuracies across 14 different tumor types and identified differentially expressed genes.
Main Results:
- The SOM analysis successfully separated tumor samples from normal expression datasets.
- Achieved classification accuracies of approximately 80% across 14 tumor types, with high accuracy for leukemia, central nervous system, melanoma, uterine, and lymphoma.
- Identified specific gene subsets with potential as tumor markers, validated against existing literature.
Conclusions:
- The developed SOM strategy offers a viable approach for unsupervised tumor classification based on gene expression data.
- The method's performance varies across tumor types, highlighting the need for high-quality data and further refinement.
- The study identified potential tumor-specific gene markers and discussed the practical utility and data requirements for this classification method.
