A Computational Framework for Tracing the Origins of Genomic Islands in Prokaryotes

Peng Wan, Dongsheng CheInternational Scholarly Research Notices 2014 · DOI

TL;DR

This paper introduces a novel computational framework to pinpoint the origins of "genomic islands" (GIs) – chunks of DNA transferred between unrelated organisms. By analyzing GIs instead of individual genes, their method builds more accurate donor-recipient networks, offering new insights into microbial evolution and adaptation.

What Problem Does This Paper Address?

Horizontal Gene Transfer (HGT) is a major driver of evolution, allowing organisms to rapidly acquire new traits. Genomic Islands (GIs) are significant HGT events, often carrying genes that confer advantages like antibiotic resistance or pathogenicity. While HGT has been studied, accurately identifying the "donor" organism for these GIs has been challenging. Previous methods often focused on individual genes, which can be less informative than analyzing the entire GI. This paper aims to bridge this gap by developing a framework that uses GIs to trace their origins and construct more robust donor-recipient networks. Understanding these relationships is crucial for fields ranging from evolutionary biology to understanding the spread of virulence factors.

Methods

The authors developed a five-step computational framework:

  1. GI Identification in Query Genome: They use SIGI-HMM, a sequence composition-based tool known for its high precision, to identify GIs within a target (query) genome.
  2. Candidate Donor Identification (BLAST): The framework then uses BLAST (specifically blastn) to search a large database of prokaryotic genomes for sequences similar to the identified GIs. These are considered initial "candidate donors."
  3. GI Prediction in Candidate Donors (AlienHunter): To filter out false positives, AlienHunter, a tool with a high recall rate for GI detection, is used to identify GIs within these candidate donor genomes.
  4. Overlapping and Filtering: The core of the filtering process involves comparing the GIs found in the query genome with the GIs found in the candidate donors. If a genomic fragment similar to a query GI is also identified as a GI in a candidate donor, that candidate is deemed a "recipient" or not a true donor for that specific GI. Only candidates whose similar fragments are not GIs in their own genomes are considered final donors.
  5. Visualization: The final donor-recipient relationships are visualized using Gephi, creating weighted networks where edge weights represent the number of GIs transferred.

Datasets

The primary dataset used is the entire collection of publicly available prokaryotic nucleotide sequences from GenBank, comprising approximately 162 million sequences and 150 billion nucleotides. This vast dataset serves as the source for identifying potential donor genomes.

Key Insights and Ideas

  • GIs as a better unit than genes: The paper argues that analyzing GIs, rather than individual genes, provides a more accurate picture of HGT events and donor-recipient relationships. This is because HGT often occurs as a cluster of sequential genes.
  • Conservative donor prediction: By requiring that a potential donor's similar genomic fragment is not itself an identified GI, the framework adopts a conservative approach, aiming to minimize false positives and identify more reliable donors.
  • Network visualization reveals patterns: Using Gephi to visualize donor-recipient networks allows for the identification of highly connected nodes (donors) and can reveal biases or preferences in HGT.
  • Potential for new discoveries: The framework successfully identified known donors consistent with previous studies, but also highlighted novel potential donors, such as Bifidobacterium animalis for Mycobacterium tuberculosis, suggesting new avenues for research.

Key Results

The framework was applied to three case studies: Mycobacterium tuberculosis H37Rv, Herminiimonas arsenicoxydans, and three Thermoanaerobacter species.

  • For M. tuberculosis H37Rv, the predicted donors included genera like Gordonia, Nocardia, and Rhodococcus, which were consistent with prior research. A novel potential donor, Bifidobacterium animalis, was also identified.
  • For H. arsenicoxydans, the analysis pointed to species within the Burkholderiales and Pseudomonadales orders as major contributors of GIs, aligning with previous findings. Janthinobacterium Marseille was identified as a highly frequent donor.
  • For the Thermoanaerobacter species, the framework extended previous network analyses, identifying T. brockii finnii Ako-1 as a significant donor contributing heat-resistant related genes.

The authors note that the computational time can be significant, with a single genome taking about an hour for GI search, and a full donor search potentially taking a day or night.

Limitations and Future Work

  • Dependence on other tools: The accuracy of the framework is inherently dependent on the performance of the underlying GI detection tools (SIGI-HMM, AlienHunter) and the BLAST search parameters (e-value cutoff).
  • Computational cost: The current framework is computationally intensive, limiting its application to a large number of genomes simultaneously.
  • Future Directions: The authors plan to scale up their framework to analyze all sequenced bacterial and archaeal genomes using clusters of computers. This will enable the construction of a large-scale, panoramic donor-recipient network to provide deeper insights into microbial gene transfer and evolution.

Summary updated

Abstract

Genomic islands (GIs) are chunks of genomic fragments that are acquired from nongenealogical organisms through horizontal gene transfer (HGT). Current researches on studying donor-recipient relationships for HGT are limited at a gene level. As more GIs have been identified and verified, the way of studying donor-recipient relationships can be better modeled by using GIs rather than individual genes. In this paper, we report the development of a computational framework for detecting origins of GIs. The main idea of our computational framework is to identify GIs in a query genome, search candidate genomes that contain genomic regions similar to those GIs in the query genome by BLAST search, and then filter out some candidate genomes if those similar genomic regions are also alien (detected by GI detection tools). We have applied our framework in finding the GI origins for Mycobacterium tuberculosis H37Rv, Herminiimonas arsenicoxydans, and three Thermoanaerobacter species. The predicted results were used to establish the donor-recipient network relationships and visualized by Gephi. Our studies have shown that donor genomes detected by our computational approach were mainly consistent with previous studies. Our framework was implemented with Perl and executed on Windows operating system.
BibTeX
@inproceedings{wan2014computational,
  author = {Peng Wan and Dongsheng Che},
  title = {{A Computational Framework for Tracing the Origins of Genomic Islands in Prokaryotes}},
  booktitle = {International Scholarly Research Notices},
  year = {2014},
  doi = {10.1155/2014/732857},
}
Figure