Gene clustering using gene expression data and self-organizing map (SOM)

Abstract

This paper presents the results of a study developing gene clustering of cancer patient.s data using gene expression data and Self-Organizing Maps (SOM). The SOM used in this paper was especially designed for patients with myelodysplastic syndromes and leukemia to reveal and present expression pattern for following disease progression since the disease itself is very aggressive. Developed SOM is self-trained using BloodSpot: a database of gene expression profiles for healthy and malignant hematopoiesis. Implemented system used expression data for IL3RA gene of cancer patients and healthy individuals containing 754 samples each. Developed SOM was made using 10×10 grid, thus 100 neurons, resulting in 100 outputs grouped in two differentiable clusters. The SOM was trained and validated using expression of NSMAF gene for both groups of individuals. Results obtained showed two successful clusters; one representing acute myeloid leukemia (AML) patients and the other representing healthy individuals. Using SOM maps the different expression patterns can be easily followed as well as used for finding a research factor for the specific disease.