5  Whole-Genome Comparison

In this part of the workshop, you will use public resources for whole-genome identification and classification of prokaryotes to help identify your isolate.

Important

Please ensure that you have downloaded the genome file for your isolate (isolate_genome.fasta) to a suitable location on your computer.

5.1 The Type (Strain) Genome Server (TYGS)

The Type (Strain) Genome Server (TYGS) is provided by the Liebniz Institute DSMZ, the German collection of Microbes and Cell Cultures. It provides genome-based classification of organisms by comparison against a reference database of over 20,000 sequenced bacterial type strains. TYGS is tightly-coupled to the authoritative LPSN (List of Prokaryotic names with Standing in Nomenclature) database, to ensure nomenclaturally-correct identification

Type Strains

A type strain is the representative of a taxon: the nomenclatural “type” on which the definition of a taxon (such as species) is based. Type strains must be deposited in at least two separate public culture collections, in different countries.

Not all taxa (in particular Candidate phyla) have type strains. It is a requirement for type strains that the organism is grown and cultured, as it cannot otherwise be placed in a culture collection. But it has been estimated that not only are the majority of bacteria and archaea so far uncultured, but they may even be unculturable in the laboratory (Hofer (2018)).

How TYGS Classification Works

TYGS takes a prokaryotic genome as input, and assigns a taxonomic identity after a series of pairwise genome comparisons.

First, the input genome is compared to the type strain database using an approach called MASH (Ondov et al. (2016)). The ten type strain genomes, and the ten most closely related type strains identified by 16S rDNA gene (extracted automatically from the input) similarity.

This set of 20 type strain genomes is then used to find the best 50 matching type strains. A phylogenetic tree is then constructed using the Genome BLAST Distance Phylogeny approach (GBDP, Henz et al. (2005)), and the resulting distances used to determine the 10 closest type strain genomes for each of the user genomes. The between-genome distances for this are calculated using digital DNA-DNA hybridisation (dDDH, Meier-Kolthoff et al. (2013)).

TYGS compiles classification output into a PDF file, and presents the resulting phylogeny on the TYGS website. All results tables and trees can be downloaded in shareable formats.

• Go to the TYGS server

• Click on Submit your query (top menu), or Submit your job (button) to reach the query page

• Click on the Browse… button and navigate to your isolate_genome.fasta file to select it for classification.
• Enter your email address in the Provide contact details field

• Click on Submit query and wait for the results

Warning

This may take some time (maybe over an hour, depending on server load!), and may not be complete before the end of the workshop.

Please move on to the next section while you are waiting.

• When you get the result confirmation email, download and inspect the result PDF, and examine the whole-genome tree.

Questions

1. What is the predicted taxonomic classification of your isolate’s genome?
2. How similar is your isolate’s genome to the closest match? (the \(d_4\) result is the dDDH (digital DNA-DNA hybridisation) score)
3. What are the most closely-related species and genera in the whole-genome tree? Is the distribution of genera consistent with what was previously known about Ochrobactrum and Brucella taxonomy?
4. What is your current opinion about the identity of your isolate? How confident are you in the identification?

5.2 genomeRxiv

genomeRxiv is a recently-developed approach to bacterial classification that promises to identify and classify prokaryotic genomes quickly and accurately into categories called LINgroups (LIN: Life Identification Number). LINgroups are a taxonomy-independent, quantitative categorisation scheme that organises genome sequences by similarity in multidimensional “space”. These categorisations can then be used to relate alternative taxonomic assignments, and other annotations, to each other (Pritchard et al. (2022)).

How genomeRxiv works

genomeRxiv works in a similar way to map grid references.

Example map grid references: point A has grid reference 970 (Easting) and 280 (Northing) as x-y co-ordinates, so the total reference is 970280. Similarly, point B is at 989319, and point C at 005255. All points in the map, however large or small, are uniquely assigned to a discrete location and, by subdividing the map into smaller and smaller squares (longer and longer numbers) an arbitrary level of precision can be reached

genomeRxiv compares input sequences to a reference database with a very fast bioinformatics algorithm (sourmash) to get a set of good matches, and then refines the match with a more precise but slower algorithm (ANI). genomeRxiv then assigns a LINgroup to the genome (Tian et al. (2021)). The LINgroup is a string of numbers, analogous to a map co-ordinate. The shorter the LINgroup, the lower the resolution of identification (Phylum, Family, etc.), and the longer the LINgroup, the finer the resolution (species, subspecies, strain, etc.).

The first key difference between LINgroups and map co-ordinates is that LINgroups are not co-ordinates on a two-dimensional surface, but in many-dimensional space. The second is that a single number describes the location, rather than two numbers (the “Easting” and “Northing” of map co-ordinates. Finally, grid references represent a physical space (such as the surface of the Earth), and LINgroups represent “sequence space” - which is not physical (Mazloom et al. (2022)).

Example arbitrary projection of 380 Ralstonia solanacearum genomes into three dimensions, based on overall genome similarity. Each sphere is a single genome. The genomes can be clustered into groups of genomes that are more similar to each other than they are to other genomes in the dataset. These clusters are assigned different colours. Each genome, such as the one indicated, has a unique co-ordinate in this “sequence space.”

Taxa are then defined by the volumes in space circumscribed by genomes which are examples of each taxon. When a new unknown genome is added, a LINgroup is assigned and - if it lies within a volume of space contained only by members of a single taxon (e.g. E. coli), the genome is assigned that taxon.

The genomeRxiv webservice allows users to input their bacterial genomes and rapidly obtain, or predict, taxonomic assignments on the basis of genome sequence.

Warning

genomeRxiv remains under active development and is not yet fully-released, although it is public and usable.

• Go to the genomeRxiv server

- Click on Identify using a FASTA file

• Either click on the Sequence to be identified link to bring up a dialogue box through which you can upload the isolate_genome.fasta file, or drag the isolate_genome.fasta file onto the text.
• Click on Identify and wait for the results

Note

The genomeRxiv site may ask you to allow pop-ups. You should allow the pop-ups, as these contain the identification output you need.

Interpreting genomeRxiv results

genomeRxiv provides three classifications:

• Tentative LIN: this is the LIN specific to the submitted genome. It may or may not match an existing, previously-assigned LIN.
• Closest Genome: this is the LIN corresponding to the closest-matching genome in the genomeRxiv database. It is unlikely to match the LIN of the submitted genome exactly, even if it’s a genome from the same species.
• Member LINgroups: this is the LINgroup that circumscribes a known taxon, as close to and enclosing the submitted genome, as possible. This is the likely identity of your isolate.

Questions

1. What is the predicted taxonomic classification of your isolate’s genome? (Check the Member LINgroups result)
2. What is the taxonomic identity assigned to the most similar genome in the database?
3. How similar is your isolate’s genome to the closest match? (Look at the ANI to Target value)
4. What is your current opinion about the identity of your isolate? Have you modified your classification? How confident are you in the identification?

Consider your final evaluation

You have now used a number of different online bioinformatics tools to obtain a possible taxonomic classification for the isolate in your blood sample. You should, by this point, have some idea of what you think the organism is, and how confident you are. Now it’s time to take a look at the official prokaryotic nomenclature database, to get a little more context around the candidate taxonomic name. Click on the link to LPSN (here, or below), to keep going.

Henz, Stefan R, Daniel H Huson, Alexander F Auch, Kay Nieselt-Struwe, and Stephan C Schuster. 2005. “Whole-Genome Prokaryotic Phylogeny.” Bioinformatics 21 (10): 2329–35.

Hofer, Ursula. 2018. “The Majority Is Uncultured.” Nature Reviews Microbiology 16 (12): 716–17. https://doi.org/10.1038/s41579-018-0097-x.

Mazloom, Reza, Leighton Pritchard, C Titus Brown, Boris A Vinatzer, and Lenwood S Heath. 2022. “LINgroups as a Principled Approach to Compare and Integrate Multiple Bacterial Taxonomies.” In Proceedings of the 13th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics. New York, NY, USA: ACM.

Meier-Kolthoff, Jan P, Alexander F Auch, Hans-Peter Klenk, and Markus Göker. 2013. “Genome Sequence-Based Species Delimitation with Confidence Intervals and Improved Distance Functions.” BMC Bioinformatics 14 (1): 60.

Ondov, Brian D., Todd J. Treangen, Páll Melsted, Adam B. Mallonee, Nicholas H. Bergman, Sergey Koren, and Adam M. Phillippy. 2016. “Mash: Fast Genome and Metagenome Distance Estimation Using MinHash.” Genome Biology 17 (1). https://doi.org/10.1186/s13059-016-0997-x.

Pritchard, Leighton, C Titus Brown, Bailey Harrington, Lenwood S Heath, N Tessa Pierce-Ward, and Boris A Vinatzer. 2022. “Could a Focus on the ‘Why’ of Taxonomy Help Taxonomy Better Respond to the Needs of Science and Society?” Front. Microbiol. 13 (May): 887310.

Tian, Long, Reza Mazloom, Lenwood S Heath, and Boris A Vinatzer. 2021. “LINflow: A Computational Pipeline That Combines an Alignment-Free with an Alignment-Based Method to Accelerate Generation of Similarity Matrices for Prokaryotic Genomes.” PeerJ 9 (e10906): e10906.