PastML

PastML infers ancestral character states on phylogenetic trees using decision-theory-based methods to assign sets of likely states to nodes and produce interpretable ancestral scenario reconstructions.

Key Features:

• Inference Methodology: Uses decision-theory concepts and the Brier score to balance marginal posterior probabilities and joint reconstructions, assigning single states in low-uncertainty regions and sets of states in higher-uncertainty regions.
• Visualization: Clusters neighboring nodes with similar state assignments and generates graph-based visualizations, including zoomable HTML maps, to depict certain and uncertain ancestral predictions.
• Speed and Efficiency: Performs rapid analysis of large phylogenetic datasets, enabling processing within minutes.

Scientific Applications:

• Phylogeography: In Dengue serotype 2 (DENV2) studies, identified main transmission routes and highlighted uncertainty in geographic origin between human and sylvatic DENV2 populations.
• Evolutionary Biology: In HIV drug-resistance analyses, revealed independent emergence of resistance mutations under treatment pressure and identified clusters of resistance corresponding to transmissions among untreated patients.

Methodology:

Applies a decision-theory framework using the Brier score to associate nodes with sets of likely states and clusters neighboring nodes for graph-based (zoomable HTML) visualization.