Hybrid fuzzy cluster ensemble framework for tumor clustering from biomolecular data

Area of Science:

• Bioinformatics and Computational Biology
• Machine Learning in Healthcare
• Cancer Genomics

Background:

• Cancer class discovery from biomolecular data is crucial for diagnosis and treatment.
• Current tumor clustering methods using gene expression data often lack robustness, stability, and accuracy.
• Single-clustering algorithms are insufficient for reliable cancer subtyping.

Purpose of the Study:

• To enhance the performance of tumor clustering from biomolecular data.
• To introduce fuzzy theory into cluster ensemble frameworks for cancer class discovery.
• To propose and evaluate four hybrid fuzzy cluster ensemble frameworks (HFCEF-I, HFCEF-II, HFCEF-III, HFCEF-IV) for identifying different cancer types.

Main Methods:

• Developed four hybrid fuzzy cluster ensemble frameworks (HFCEF-I to HFCEF-IV).
• HFCEF-I and HFCEF-II utilize Affinity Propagation (AP) with different clustering dimensions (sample vs. attribute) and fuzzy logic.
• HFCEF-III and HFCEF-IV combine HFCEF-I and HFCEF-II serially and concurrently, respectively, using consensus functions like fuzzy c-means or Normalized Cut (Ncut).

Main Results:

• The proposed hybrid fuzzy cluster ensemble frameworks demonstrate strong performance on real-world datasets, particularly biomolecular data.
• These novel frameworks significantly improve robustness, stability, and accuracy in tumor clustering.
• Experimental results show superior performance compared to state-of-the-art single clustering algorithms and traditional cluster ensemble methods.

Conclusions:

• Hybrid fuzzy cluster ensemble frameworks offer a more effective approach to cancer class discovery.
• The proposed methods provide a robust and accurate solution for tumor subtyping using gene expression data.
• This research advances the application of machine learning in precision oncology.