A NEW CLUSTERING METHOD AND ITS APPLICATION TO PROTEOMIC PROFILING FOR COLON CANCER

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Clustering methods are crucial for analyzing complex biological datasets.
• Existing methods often produce large numbers of clusters and do not handle overlapping clusters well.
• The NCI-60 dataset provides a valuable resource for cancer cell line research.

Purpose of the Study:

• To introduce a new clustering method, quasi-clique merger, designed for biological data analysis.
• To develop associated data pretreatment programs for enhanced clustering.
• To evaluate the method's performance on cancer cell line data and identify potential diagnostic markers.

Main Methods:

• Development of the quasi-clique merger algorithm for non-binary hierarchical clustering.
• Implementation of data pretreatment programs to support the clustering method.
• Application of the method to cluster the NCI-60 human cancer cell lines using proteomic determinants for 5-Fluorouracil (5-FU) chemosensitivity.

Main Results:

• The quasi-clique merger method generated a significantly smaller number of clusters compared to previous methods.
• Overlapping clusters were successfully produced, allowing for nuanced data representation.
• All colon cancer cell lines were grouped into a single cluster, highlighting eight proteomic markers as potential colon cancer diagnostic indicators.
• The new method demonstrated superior performance over existing approaches on the NCI-60 dataset.

Conclusions:

• The quasi-clique merger method offers an effective approach for clustering biological data, particularly for cancer cell line analysis.
• The identified proteomic markers show promise as diagnostic tools for colon cancer.
• This novel clustering technique advances the analysis of complex biological datasets and biomarker discovery.