EPA-ng: Massively Parallel Evolutionary Placement of Genetic Sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Next-generation sequencing (NGS) generates vast amounts of molecular sequence data, posing challenges for analysis, especially in metagenetics.
• Phylogenetic placement is crucial for understanding the evolutionary context of sequences from diverse microbial environments.
• Existing phylogenetic placement algorithms like RAxML-EPA and PPLACER face scalability limitations with increasing NGS data volumes.

Purpose of the Study:

• To present EPA-NG, a reimplementation of the evolutionary placement algorithm (EPA).
• To address the scalability limitations of current phylogenetic placement tools for large metagenetic datasets.
• To offer a faster and more parallelizable solution for taxonomic identification in microbial ecology.

Main Methods:

• EPA-NG is a complete reimplementation of the EPA, integrating features from RAxML-EPA and PPLACER.
• It supports both shared memory and distributed memory parallelization, enabling execution on computing clusters.
• Performance was assessed by placing 1 billion metagenetic reads from the Tara Oceans Project onto a reference tree.

Main Results:

• EPA-NG demonstrates substantial speed improvements, outperforming RAxML-EPA and PPLACER by up to 30x in sequential mode.
• The tool achieves comparable parallel efficiency on shared memory systems.
• Distributed memory parallelization of EPA-NG shows excellent scalability, performing well up to 2048 cores.

Conclusions:

• EPA-NG offers a significantly faster and more scalable solution for phylogenetic placement of NGS data.
• The tool effectively handles large-scale metagenetic datasets, advancing microbial ecology research.
• EPA-NG is available for broader scientific use, facilitating more comprehensive evolutionary analyses.