mtDeOxoGer: A machine learning-based strategy to effectively filter artifactual mutations induced by oxidative damage...

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Mitochondrial DNA (mtDNA) mutations are common in tumors, with next-generation sequencing (NGS) enhancing detection.
• Oxidative artifacts, especially from ultrasonic DNA fragmentation, compromise mtDNA NGS accuracy by mimicking low-frequency mutations.
• The impact of oxidative damage on mtDNA sequencing is less understood than in nuclear DNA.

Purpose of the Study:

• To systematically evaluate oxidative damage artifacts in mtDNA NGS data.
• To develop a machine learning strategy for improving the accuracy of mtDNA mutation detection.
• To identify the sources of oxidative artifacts in mtDNA library preparation.

Main Methods:

• Analysis of private and public mtDNA NGS datasets, focusing on 8-oxoguanine lesions.
• Experimental validation of ultrasonic DNA fragmentation parameters and enzyme-based fragmentation.
• Development and application of a logistic regression model (mtDeOxoGer) incorporating variant allele frequency, strand orientation bias, GC content, and sequence context.

Main Results:

• Low-frequency C>A and G>T substitutions were identified as common artifacts in mtDNA NGS.
• Ultrasonic fragmentation was confirmed as a major source of oxidative artifacts, influenced by temperature, intensity, and cycle number.
• The mtDeOxoGer model effectively distinguished authentic mtDNA mutations from oxidative artifacts.

Conclusions:

• Refining mtDNA sequencing methodologies is crucial for accurate genetic studies.
• The developed machine learning strategy offers a practical framework for enhancing the reliability of mtDNA mutation detection.
• This approach supports more dependable mitochondrial genetics research in cancer and other diseases.