Gretl-variation GRaph Evaluation TooLkit

Abstract

Motivation: As genome graphs are powerful data structures for representing the genetic diversity within populations, they can help identify genomic variations that traditional linear references miss, but their complexity and size makes the analysis of genome graphs challenging. We sought to develop a genome graph analysis tool that helps these analyses to become more accessible by addressing the limitations of existing tools. Specifically, we improve scalability and user-friendliness, and we provide many new statistics tailored to variation graphs for graph evaluation, including sample-specific features.

Results: We developed an efficient, comprehensive, and integrated tool, gretl, to analyze genome graphs and gain insights into their structure and composition by providing a wide range of statistics. gretl can be utilized to evaluate different graphs, compare the output of graph construction pipelines with different parameters, as well as perform an in-depth analysis of individual graphs, including sample-specific analysis. With the assistance of gretl, novel patterns of genetic variation and potential regions of interest can be identified, for later, more detailed inspection. We demonstrate that gretl outperforms other tools in terms of speed, particularly for larger genome graphs.

Availability and implementation: Commented Rust source code and documentation is available under MIT license at https://github.com/MoinSebi/gretl together with Python scripts and step-by-step usage examples. The package is available at Bioconda for easy installation.

Figure 1.

Gretl overview. (A) Genome graph construction workflow: genome graph properties are influenced by various factors, including parameter selection, sample curation, and methodology, all of which impact the layout and structure of the resulting genome graph. For evaluation purposes, multiple graphs can be simultaneously generated and compared to identify an optimal representation for a specific task. The selected graph can then be analyzed with gretl. (B) Visualization of gretl output: left, graphs can be clustered based on multiple statistics, grouping similar species or construction parameters (shown here, with normalized values). Right, scatter plot depicting two selected statistics across various graphs, facilitating comparisons between different species. (C) In-depth analysis of a selected genome graph (example from yeast): left, path-centric sliding window analysis of the Saccharomyces cerevisiae genome graph, highlighting regions of high similarity. Right, pan-genomic analysis of the genome graph. Sequences found only in a single sample are separated and each block represents one path of the graph.