EvANI benchmarking workflow for evolutionary distance estimation.

Area of Science:

• Genomics
• Bioinformatics
• Evolutionary Biology

Background:

• Long-read sequencing advances enable high-quality genome assemblies, facilitating evolutionary relationship studies.
• Average nucleotide identity (ANI) is crucial for species delineation and database searching but computationally intensive.
• Sketch-based methods accelerate ANI estimation using heuristics, necessitating rigorous evaluation.

Purpose of the Study:

• To introduce EvANI, an evaluation framework for assessing genome comparison algorithms.
• To analyze the impact of assumptions and heuristics in sketch-based ANI estimation.
• To identify optimal methods for accurate and efficient genome-wide comparisons.

Main Methods:

• Development of simulated and real benchmark datasets.
• Application of a rank-correlation-based metric for evaluating distance estimates.
• Comparison of various ANI estimation algorithms, including ANIb and k-mer based approaches.

Main Results:

• ANIb provides the most accurate tree distance but is computationally inefficient.
• K-mer based methods demonstrate high efficiency and consistent accuracy in ANI estimation.
• Optimal k-mer lengths vary across clades, suggesting a need for multiple k values (e.g., k=10, k=19 for Chlamydiales).
• Maximal exact match approaches offer a balance between efficiency and accuracy.

Conclusions:

• EvANI provides a robust method for evaluating genome comparison tools.
• K-mer based methods are highly effective for large-scale genomic comparisons.
• A single fixed k-mer length may not be universally optimal; adaptable approaches are beneficial.