Maximum likelihood pandemic-scale phylogenetics.

Area of Science:

• Genomics
• Computational Biology
• Epidemiology

Background:

• Phylogenetic analysis is vital for interpreting genomic data, particularly for tracking viral spread and evolution, as demonstrated by SARS-CoV-2 studies.
• Current model-based phylogenetic methods, while accurate, struggle to scale with the massive genomic datasets generated during pandemics.
• This computational limitation hinders a comprehensive understanding of viral evolution and transmission dynamics.

Approach:

• Developed novel computational methods by reworking Felsenstein's pruning algorithm for scalable, likelihood-based phylogenetic analysis of large genomic datasets.
• Leveraged near-certainty of ancestral genomes and similarities in closely related, densely sampled genomes to significantly reduce computational resource demands.
• Integrated new tree-searching techniques, resulting in the MAPLE (MAximum Parsimonious Likelihood Estimation) software.

Key Points:

• MAPLE software outperforms popular phylogenetic tools like FastTree 2, IQ-TREE 2, RAxML-NG, and UShER in accuracy and efficiency.
• The new approach enables complex and accurate probabilistic phylogenetic analyses on millions of microbial genomes.
• Significantly reduces computational time and memory requirements for phylogenetic inference.

Conclusions:

• The developed methods and MAPLE software enable complex, accurate probabilistic phylogenetic analyses on millions of microbial genomes.
• This advancement extends the capabilities of genomic epidemiology, allowing for deeper insights into pathogen evolution and spread.
• The scalable approach is crucial for future, even larger, epidemiological datasets and other fields like metagenomics and biodiversity science.