Modeling of Cancer Classifier to Predict Site of Origin

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer classification by origin is crucial for effective treatment and prediction.
• Existing methods, such as nucleotide-based analysis, have limitations in accuracy and sensitivity.

Purpose of the Study:

• To develop a novel cancer gene classification method based on amino acid sequences.
• To improve the accuracy and sensitivity of predicting cancer origin.

Main Methods:

• Representing cancer gene networks as equivalent electrical circuits using amino acid properties (hydrophilic/hydrophobic).
• Modeling a classifier analyzing Nyquist curve characteristics (phase value, peak gain, shape).
• Testing the model on 93 Homo Sapiens cancer gene samples.

Main Results:

• Achieved 81.09% classification accuracy for cancer gene origin.
• Demonstrated 69% better performance compared to existing nucleotide-based methods.
• The model proved sensitive and simple for cancer classification.

Conclusions:

• The proposed electrical circuit model effectively classifies cancer genes based on their site of origin.
• This approach offers a significant advancement over traditional nucleotide-based methods.
• The model's sensitivity and simplicity make it a valuable tool for cancer research.