LYCEUM: learning to call copy number variants on low-coverage ancient genomes

Area of Science:

• Genomics
• Bioinformatics
• Evolutionary Biology

Background:

• Copy number variants (CNVs) are key drivers of phenotypic variation and adaptation.
• CNVs contribute to various genetic disorders, making ancient DNA (aDNA) critical for understanding disease susceptibility.
• Detecting CNVs in degraded, contaminated, and low-coverage aDNA is challenging for conventional algorithms.

Purpose of the Study:

• To introduce LYCEUM, the first machine learning-based CNV caller specifically designed for ancient DNA.
• To address the limitations of existing CNV detection methods in low-quality and low-coverage aDNA data.
• To enable accurate CNV identification in ancient genomes for evolutionary and disease studies.

Main Methods:

• Developed LYCEUM, a novel machine learning model for aDNA CNV detection.
• Employed a two-step training strategy: pre-training on high-coverage data (1000 Genomes Project) and fine-tuning on limited high-confidence aDNA CNV calls.
• Adapted the model to accurately call CNVs from downsampled read-depth signals characteristic of aDNA.

Main Results:

• LYCEUM achieves accurate CNV detection in low-coverage ancient genomes.
• Segmental deletion calls by LYCEUM correlate with sample demographic history.
• CNV patterns identified by LYCEUM show evidence of natural selection.

Conclusions:

• LYCEUM overcomes significant challenges in aDNA CNV detection.
• The tool facilitates deeper insights into the role of CNVs in ancient population adaptation and evolution.
• LYCEUM provides a robust method for analyzing structural variations in ancient human and non-human genomes.