New method finds areas of high change in bacterial genomes
Researchers developed a framework to identify hotspots in bacterial genomes using simulation-based estimation.
Scientists have created a new framework to find specific areas in bacterial genomes where mutation and recombination rates are high. These areas are called hotspots. These hotspots often show where selection acts on certain genes.
Detecting these hotspots is difficult because mutation and recombination leave similar signals in sequence data. Clonal relationships also act as a confounding factor. Previous methods often assumed mutation rates stayed constant or used only a few summary statistics.
The new method uses a simulation-based approach. It relies on a model called ClonalOrigin. This model is an approximation of the ancestral recombination graph with gene conversion. It allows researchers to control the clonal genealogy.
The researchers used neural posterior estimation, or NPE, to perform the inference. They trained the NPE on simulations. This training included a wide range of summary statistics. The model also used gene length as both an observed statistic and a parameter.
To find hotspots, the team used a circular posterior-propagated Difference-of-Gaussians detector. This tool compares local and regional smoothers of the posterior log rates. It accounts for the circular structure of the genome and includes uncertainty in hotspot probabilities.
In tests using simulated data, the method recovered the true rates. The credible intervals became narrower as the length of the gene increased. The researchers applied the method to real datasets from Bacillus cereus, Staphylococcus aureus, and Escherichia coli.
Recombination in bacteria occurs when a small fragment of a donor genome integrates into a recipient genome. This process is similar to gene conversion in eukaryotes. The ancestral process is described by an ancestral recombination graph with gene conversion.
The ClonalOrigin model assumes recombination events happen independently along the clonal genealogy. Non-clonal lineages can only coalesce with the clonal genealogy. This restriction helps reduce the computational cost of simulations.
Key facts
- The method uses neural posterior estimation to infer gene-specific rates.
- The framework is applicable to different gene lengths by learning their effects.
- The tool was tested on Bacillus cereus, Staphylococcus aureus, and Escherichia coli.
- The implementation is available in Python at https://github.com/haochu-liu/bacnpe.