A Kernelized Classification Approach for Cancer Recognition Using Markovian Analysis of DNA Structure Patterns as...

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• DNA and RNA are crucial nucleotide-based molecules for cellular processes, containing genetic material that dictates cell growth and function.
• Specific DNA structure patterns influence cell behavior and physiological effects, making their analysis vital for understanding cancer and other disorders.
• Identifying these patterns is key for early disease detection, significantly impacting cancer research and treatment efficacy.

Purpose of the Study:

• To investigate dinucleotide structure patterns across various genomic regions: non-coding region sequence (N-CDS), coding region sequence (CDS), and whole raw DNA sequence (W.R. sequence).
• To analyze dinucleotide patterns in both malignant and non-malignant DNA sequences within these diverse genetic environments.
• To evaluate the efficacy of Markovian modeling for predicting dinucleotide probabilities, aiming to reduce feature complexity and computational costs.

Main Methods:

• Utilized Markovian modeling to predict dinucleotide probabilities, offering a computationally efficient alternative to Kernelized Logistic Regression (KLR) and Support Vector Machine (SVM).
• Examined dinucleotide patterns in whole raw DNA sequences (W.R. sequence), coding region sequences (CDS), and non-coding region sequences (N-CDS).
• Validated the approach using accuracy metrics and 10-fold cross-validation in case studies.

Main Results:

• Markovian modeling effectively reduced feature complexity and computational costs compared to KLR and SVM.
• The combination of Markovian probability-generated classifiers and feature reduction demonstrated strong performance.
• The study successfully distinguished between DNA sequences associated with cancer and those from non-cancerous diseases.

Conclusions:

• Dinucleotide pattern analysis in genomic regions, including CDS and N-CDS, provides a robust method for cancer detection.
• Markovian modeling offers an efficient computational approach for analyzing DNA sequences and predicting disease association.
• The findings support the potential of this approach for improving early cancer diagnostics and research.