Chronumental

Chronumental converts phylogenetic divergence-trees into time trees where branch lengths represent temporal intervals, enabling estimation of node dates to study evolutionary dynamics of rapidly evolving organisms such as viruses.

Key Features:

• Time-tree conversion: Transforms divergence-based phylogenies so distances represent temporal intervals rather than genetic substitutions.
• Node date estimation: Estimates chronological dates for nodes on a phylogeny, including when metadata for some nodes are unavailable.
• Probabilistic framework: Fits branch lengths within a probabilistic model to represent uncertainty in temporal inference.
• Optimization algorithm: Uses stochastic gradient descent to optimize branch lengths.
• Objective function: Maximizes the evidence lower bound (ELBO) during model optimization.
• Scalability: Designed to infer time trees from phylogenies containing millions of nodes.
• XLA acceleration: Leverages XLA (Accelerated Linear Algebra) compilation frameworks to accelerate computation.
• Rapid computation: Produces chronological predictions on large datasets in a short time (reported in minutes).

Scientific Applications:

• Phylogenetic dating: Infers temporal relationships among nodes for time-resolved phylogenies.
• Viral evolution analysis: Supports study of evolutionary dynamics of rapidly evolving viruses, including SARS-CoV-2.
• Large-scale temporal analyses: Enables analysis of temporal aspects of genetic divergence across very large sequencing datasets.

Methodology:

Chronumental optimizes branch lengths using stochastic gradient descent within a probabilistic framework by maximizing the evidence lower bound, and employs XLA (Accelerated Linear Algebra) compilation to accelerate computation on phylogenies of millions of nodes.