Data mining on DNA sequences of hepatitis B virus

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Hepatitis B Virus (HBV) infection is a major cause of liver cancer (HCC).
• Identifying genomic markers associated with HCC development from HBV sequences is crucial for early diagnosis and treatment.
• Large experimental datasets in bioinformatics require robust data mining frameworks for meaningful information extraction.

Purpose of the Study:

• To develop and evaluate a data mining framework for identifying genomic markers in HBV associated with HCC.
• To compare the efficacy of novel classification methods (Rule Learning and Nonlinear Integral) against existing approaches.
• To validate the identified markers and classification performance on independent datasets.

Main Methods:

• A data mining framework encompassing molecular evolution analysis, clustering, feature selection (Information Gain), and classifier learning (Rule Learning, Nonlinear Integral).
• Collection and analysis of HBV DNA sequences (genotype B and C) from over 200 patients.
• Development of a clustering method to identify subgroups within HBV genotype C and a rule learning algorithm based on Evolutionary Algorithm.

Main Results:

• Three distinct subgroups were identified within HBV genotype C, with a developed clustering method for separation.
• Important mutation markers (sites) associated with HCC were identified for HBV genotype B and genotype C subgroups (C1, C2, C3).
• The developed classification methods achieved over 70% accuracy and 80% sensitivity in classifying new datasets.

Conclusions:

• The data mining framework effectively identifies genomic markers and interactions in HBV linked to HCC.
• The Rule Learning and Nonlinear Integral classifiers demonstrate high performance for initial liver cancer screening.
• These findings contribute to the early diagnosis of liver cancer by analyzing HBV genomic variations.