Pathway-based Human Disease Clustering Tool using Self-Organizing Maps

Understanding how different diseases are related to one another based on their shared pathways could provide new insights into disease etiology and classification. The exploration of disease-disease associations by using a system-level biological data is made possible as the data is now publicly available via databases such as in the database maintained by Kyoto Encyclopedia of Genes and Genomes (KEGG). By being able to cluster and visualize relationships with respect to shared entities on the pathways of human diseases, researchers would be fully able to use the available pathway databases for the said scientific purposes. Thus, there is a need for an algorithm that is able to effectively visualize the topology of a multidimensional data. Self-Organizing Map (SOM) is a type of artificial neural network that employs unsupervised learning capable of discovering patterns in datasets by reducing multidimensional data to a low-dimensional representation. SOM can be used as a pre-processing step for cluster analysis via k-means clustering and hierarchical clustering. PathSOM is a software that uses self-organizing maps to create different visualizations of the underlying relationships of human diseases utilizing the available pathway data from KEGG.