fineSTRUCTURE

fineSTRUCTURE performs model-based Bayesian clustering of individuals using coancestry matrices derived from ChromoPainter chromosome painting to infer fine-scale population structure from dense sequencing and full resequencing data.

Key Features:

• ChromoPainter summary statistic: Uses ChromoPainter output to produce a coancestry matrix that serves as a nearly sufficient summary statistic capturing haplotype sharing among individuals.
• Model-based Bayesian clustering: Applies a model-based Bayesian clustering algorithm to the coancestry matrix to infer discrete populations and quantify assignment uncertainty.
• Chromosome painting and haplotype reconstruction: Treats each individual as a recipient whose chromosomes are reconstructed from DNA segments donated by other individuals to capture patterns of haplotype similarity.
• Scalability: Designed to analyze dense sequencing and full resequencing datasets at the scale of hundreds to thousands of individuals and hundreds of thousands of markers.
• Integration with LD and marker information: The coancestry matrix integrates information across successive markers when linkage disequilibrium is present, and nearly completely encompasses information used when markers are independent.
• Comparative resolution and sensitivity: Provides increased sensitivity to subtle population structure and greater capacity to separate distinct populations relative to allele-frequency-based PCA or STRUCTURE approaches.
• Computational efficiency: Employs an inference framework intended to capture population-structure information efficiently from dense haplotype data.

Scientific Applications:

• Fine-scale population inference: Infers population structure at continental, regional, local, and family scales, as demonstrated on the Human Genome Diversity Panel.
• Large-scale resequencing analysis: Detects population clusters in full resequencing datasets comprising hundreds to thousands of individuals and hundreds of thousands of markers (e.g., 938 individuals and 641,000 markers identifying 226 populations in HGDP).
• Summary statistic for downstream methods: Provides a coancestry matrix that encapsulates information used by PCA and model-based approaches such as STRUCTURE for complementary analyses.

Methodology:

ChromoPainter performs chromosome painting by reconstructing each recipient's chromosomes from donor haplotype segments to produce a coancestry matrix, which fineSTRUCTURE analyzes using a model-based Bayesian clustering algorithm to infer populations and assignment uncertainty.