SCAMPP

SCAMPP scales likelihood-based phylogenetic placement to efficiently place query sequences into ultra-large precomputed backbone trees for large-scale phylogenetic inference and taxon identification.

Key Features:

• Scalability: Enhances scalability of likelihood-based phylogenetic placement methods such as pplacer and EPA-ng, which require the query sequence to be included within a multiple sequence alignment encompassing all backbone leaf sequences.
• Ultra-large backbone trees: Extends placement capability to backbone trees with up to 200,000 leaves, addressing previous limitations encountered for trees larger than ~50,000 leaves.
• Accuracy: Maintains high placement accuracy and outperforms fast placement methods APPLES and APPLES-2.
• Implementations: Provided as pplacer-SCAMPP and EPA-ng-SCAMPP implementations that apply the SCAMPP approach to those likelihood engines.

Scientific Applications:

• Large-scale phylogenetic inference: Enables placement of new sequences into massive backbone trees to support construction of extensive phylogenies.
• Taxon identification: Supports accurate taxon identification of newly obtained query sequences via placement into precomputed reference trees.
• Biodiversity and evolutionary studies: Facilitates analyses requiring detailed phylogenetic context across very large taxon collections.

Methodology:

SCAMPP builds on the maximum likelihood framework used by pplacer and EPA-ng and introduces computational enhancements to allow these likelihood-based placement methods to operate efficiently on ultra-large backbone trees.