ScisTree

ScisTree infers cell lineage trees and calls genotypes from noisy single-cell genotype data using individualized genotype probabilities.

Key Features:

• Individualized Genotype Probabilities: Allows specification of different probabilities for each genotype at both the cell (row) and site (column) levels to model non-uniform uncertainty.
• Support for Binary and Ternary Genotypes: Handles both binary and ternary genotype encodings.
• Infinite Sites Model Assumption: Operates under the infinite sites model, assuming no site reversion or parallel mutations.
• Fast Heuristic Algorithm: Implements a fast heuristic that maximizes genotype likelihoods while enforcing a perfect phylogeny constraint.
• Efficiency and Scalability: Demonstrated improved accuracy and computational efficiency in simulation studies and can scale to large datasets, enabling tasks such as imputation of doublets.

Scientific Applications:

• Cell lineage reconstruction: Infers cell lineage trees from noisy single-cell genotype data under the infinite sites model.
• Genotype calling: Calls genotypes from noisy single-cell genotype data using per-genotype probability inputs from single-cell genotype callers.
• Doublet imputation: Enables imputation of doublets—cells that appear as a single entity but derive from two distinct cells—within single-cell genotype datasets.
• Biological studies: Supports analyses in developmental biology, cancer genomics, and evolutionary studies that require reconstruction of cell lineage histories.

Methodology:

Integrates genotype probabilities computed by existing single-cell genotype callers into a fast heuristic algorithm that searches for genotypes maximizing likelihood while enforcing a perfect phylogeny under the infinite sites model.