A Comparative Evaluation of Cancer Classification via TP53 Gene Mutations Using Machin Learning

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer classification relies on identifying cancer-causing gene mutations.
• Current methods lack effective prediction for all cancer types.
• Early detection of mutated tumor protein P53 (TP53) can guide treatment.

Purpose of the Study:

• To predict cancer types using data mining.
• To identify effective machine learning algorithms for cancer classification.
• To analyze TP53 gene mutations for improved cancer prediction.

Main Methods:

• Utilized the Universal Mutation Database (UMD-2010) for gene mutation data.
• Applied bioinformatics techniques including pairwise alignment and BLAST.
• Employed machine learning algorithms, specifically Multi-Layer Perceptron (MLP) and Support Vector Machine (SVM), for classification.

Main Results:

• Both MLP and SVM proved effective for cancer classification.
• SVM achieved a higher accuracy of 93.7% compared to MLP's 90%.
• SVM exhibited a lower Mean Absolute Error (MAE) than MLP.

Conclusions:

• Support Vector Machine (SVM) is a more accurate algorithm than Multi-Layer Perceptron (MLP) for classifying cancer based on TP53 gene mutations.
• The developed data mining technique offers a promising approach for cancer type prediction.
• Further research can leverage these findings for enhanced cancer diagnostics and therapeutics.