A Machine-Learning-Based Approach to Prediction of Biogeographic Ancestry within Europe.

Area of Science:

• Population genetics
• Bioinformatics
• Computational biology

Background:

• Massive parallel sequencing (MPS) generates large datasets valuable for population genetics.
• Machine learning (ML) techniques are well-suited for analyzing complex MPS data.
• Slavic populations represent a significant portion of Europe's population with close genetic ties.

Purpose of the Study:

• To evaluate the feasibility of using ML to classify DNA samples from Slavic and non-Slavic individuals.
• To assess the ability to discern genetic provenance within closely related Slavic populations.
• To categorize DNA specimens from diverse Slavic population representatives.

Main Methods:

• Pre-processing of raw sequencing data into a 1200-character binary vector.
• Application of three ML classifiers: Random Forest, Support Vector Machine (SVM), and XGBoost.
• Empirical evaluation of classifier performance for genetic ancestry inference.

Main Results:

• Support Vector Machine (SVM) with a linear kernel demonstrated superior performance.
• Achieved 99.9% accuracy in classifying Slavic and non-Slavic individuals.
• Obtained high F1-scores (0.9846-1.000) across all classes, indicating robust classification.

Conclusions:

• ML, particularly SVM, is highly effective for classifying genetic ancestry in large sequencing datasets.
• This approach can accurately distinguish individuals from different Slavic populations, even those with close genetic origins.
• The study confirms the potential of MPS and ML for detailed population genetics research.