Microbiology, Metagenomics and Bioinformatics

Johan Bengtsson-Palme, University of Gothenburg

Browsing Posts tagged Species classification

Yesterday, Molecular Ecology Resources put online an unedited version of a recent paper which I co-authored. This time, Rodney Richardson at Ohio State University has made a tremendous work of evaluating three taxonomic classification software – the RDP Naïve Bayesian Classifier, RTAX and UTAX – on a set of DNA barcoding regions commonly used for plants, namely the ITS2, and the matK, rbcL, trnL and trnH genes.

In the paper (1), we discuss the results, merits and limitations of the classifiers. In brief, we found that:

  • There is a considerable trade-off between accuracy and sensitivity for the classifiers tested, which indicates a need for improved sequence classification tools (2)
  • UTAX was superior with respect to error rate, but was exceedingly stringent and thus suffered from a low assignment rate
  • The RDP Naïve Bayesian Classifier displayed high sensitivity and low error at the family and order levels, but had a genus-level error rate of 9.6 percent
  • RTAX showed high sensitivity at all taxonomic ranks, but at the same time consistently produced the high error rates
  • The choice of locus has significant effects on the classification sensitivity of all tested tools
  • All classifiers showed strong relationships between database completeness, classification sensitivity and classification accuracy

We believe that the methods of comparison we have used are simple and robust, and thereby provides a methodological and conceptual foundation for future software evaluations. On a personal note, I will thoroughly enjoy working with Rodney and Reed again; I had a great time discussing the ins and outs of taxonomic classification with them! The paper can be found here.

References and notes

  1. Richardson RT, Bengtsson-Palme J, Johnson RM: Evaluating and Optimizing the Performance of Software Commonly Used for the Taxonomic Classification of DNA Sequence Data. Molecular Ecology Resources, Early view (2016). doi: 10.1111/1755-0998.12628 [Paper link]
  2. This is something that several classifiers also showed in the evaluation we did for the Metaxa2 paper (3). Interestingly enough, Metaxa2 is better at maintaining high accuracy also as sensitivity is increased.
  3. Bengtsson-Palme J, Hartmann M, Eriksson KM, Pal C, Thorell K, Larsson DGJ, Nilsson RH: Metaxa2: Improved identification and taxonomic classification of small and large subunit rRNA in metagenomic data. Molecular Ecology Resources, 15, 6, 1403–1414 (2015). doi: 10.1111/1755-0998.12399 [Paper link]

I am very happy to announce that Metaxa2 version 2.1 has been released today. This new version brings a lot of important improvements to the Metaxa2 software (1), in particular by the introduction of the Metaxa2 Diversity Tools. This is the list of new features (further elaboration follows below):

  • The Metaxa2 Diversity Tools:
    • metaxa2_dc – a tool for collecting several .taxonomy.txt output files into one large abundance matrix, suitable for analysis in, e.g., R
    • metaxa2_rf – generates rarefaction curves based on the .taxonomy.txt output
    • metaxa2_si – species inference based on guessing species data from the other species present in the .taxonomy.txt output file
    • metaxa2_uc – a tool for determining if the community composition of a sample is significantly different from others through resampling analysis
  • Added a new detection mode for detection of multiple rRNA in the same sequence, e.g. a genome
  • Added the --reference option to improve the use of Metaxa2 as a tool to sort out host rRNA sequences from a dataset
  • Added the --split_pairs option causing Metaxa2 to output paired-end sequences into two separate files, which is nice for further analysis of rRNA reads
  • The default setting for the --align option has been changed to ‘none
  • Automatic detection of which BLAST package that is installed
  • Fixed a bug causing the last read of paired-end FASTA input to be ignored
  • Fixed an occasionally occurring BLAST+ related warning message
  • Fixed a bug that could cause the classifier to crash on highly divergent BLAST matches

The new version of Metaxa2 can be downloaded here, and for those interested I will spend the rest of this post outlining the new features.

Metaxa2 Diversity Tools
One often requested feature of Metaxa2 is the ability to further make simple analysis from the data after classification. The Metaxa2 Diversity Tools included in Metaxa2 2.1 is a seed for such an effort (although not close to a full-fledge community analysis package compared to QIIME (2) or Mothur (3)). The set currently consist of four tools

The Metaxa2 Data Collector (metaxa2_dc) is the simplest of them (but probably the most requested), designed to merge the output of several *.level_X.txt files from the Metaxa2 Taxonomic Traversal Tool into one large abundance matrix, suitable for further analysis in, for example, R. The Metaxa2 Species Inference tool (metaxa2_si) can be used to further infer taxon information on, for example, the species level at a lower reliability than what would be permitted by the Metaxa2 classifier, using a complementary algorithm. The idea is that is if only a single species is present in, e.g., a family and a read is assigned to this family, but not classified to the species level, that sequence will be inferred to the same species as the other reads, given that it has more than 97% sequence identity to its best reference match. This can be useful if the user really needs species or genus classifications but many organisms in the studied species group have similar rRNA sequences, making it hard for the Metaxa2 classifier to classify sequences to the species level.

The Metaxa2 Rarefaction analysis tool (metaxa2_rf) performs a rarefaction analysis based on the output from the Metaxa2 classifier, taking into account also the unclassified portion of rRNAs. The Metaxa2 Uniqueness of Community analyzer (metaxa2_uc), finally, allows analysis of whether the community composition of two or more samples or groups is significantly different. Using resampling of the community data, the null hypothesis that the taxonomic content of two communities is drawn from the same set of taxa (given certain abundances) is tested. All these tools are further described in the manual.

The genome mode
Metaxa2 has long been said not to be useful for predicting rRNA in longer sequences, such as full genomes or chromosomes, since it has traditionally only looked for a single rRNA hit. With Metaxa2 2.1, it is now possible to use Metaxa2 on longer sequences to detect multiple rRNA occurrences. To do this, you need to change the operating mode using the new --mode option to either ‘auto‘ or ‘genome‘. The auto mode will treat sequences longer than 2500 bp as “genome” sequences and look for multiple matches in these.

The reference mode
Another feature request that has been addressed in the new Metaxa2 version is the ability to filter out certain sequences from the data set. For example, you may want to exclude all rRNA sequences that are derived from to host organism, but keep the analysis of all other rRNA reads. This is now possible by supplying a file of reference rRNA sequences to exclude in FASTA format to the --reference option.

Experimental Usearch support
Finally, we have toyed around with support for Usearch (4) instead of BLAST (5) as the search algorithm for the classification step. However, this is far from fine-tuned and it is included as an experimental feature that you may use on your own risk! We recommend that you not use it for classification of data for publication yet. However, we are interested in how this works for you, so if you like you may test to run the Usearch algorithm in parallel with your BLAST-based analysis and compare the results and send me your input on how it works. You can read more about using Usearch at the end of the Metaxa2 manual.

References

  1. Bengtsson-Palme J, Hartmann M, Eriksson KM, Pal C, Thorell K, Larsson DGJ, Nilsson RH: Metaxa2: Improved Identification and Taxonomic Classification of Small and Large Subunit rRNA in Metagenomic Data. Molecular Ecology Resources (2015). doi: 10.1111/1755-0998.12399 [Paper link]
  2. Caporaso JG, Kuczynski J, Stombaugh J et al.: QIIME allows analysis of high-throughput community sequencing data. Nature Methods, 7, 335–336 (2010).
  3. Schloss PD, Westcott SL, Ryabin T et al.: Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Applied and Environmental Microbiology, 75, 7537–7541 (2009).
  4. Edgar RC: Search and clustering orders of magnitude faster than BLAST. Bioinformatics, 26, 2460–2461 (2010).
  5. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res, 25, 3389–3402 (1997).

Metaxa2 has been updated to version 2.0.2 and can be downloaded from the Metaxa2 web site. The 2.0.2 update fixes two minor bugs; one causing the “.graph” file to display incorrect or no names for the regions of the LSU regions, and one causing misreporting of the number of sequences in single-end FASTQ files (paired-end files were reported correctly). The update also brings a slightly improved classifier. Thanks to Marco Severgnini for reporting the FASTQ file issue! The update is available here.

After almost a year in different stages of review and revision, in which the paper (but not the software) saw a total transformation, I am happy to announce that the paper describing Metaxa2 has been accepted in Molecular Ecology Resources and is available in a rudimentary online early form. The figures in this version are not that pretty, but those who wants to read the paper asap, you have the possibility to do so.

This means that if you have been using Metaxa2 for a publication, there is now a new preferred way of citing this, namely:

Bengtsson-Palme J, Hartmann M, Eriksson KM, Pal C, Thorell K, Larsson DGJ, Nilsson RH: Metaxa2: Improved Identification and Taxonomic Classification of Small and Large Subunit rRNA in Metagenomic Data. Molecular Ecology Resources (2015). doi: 10.1111/1755-0998.12399

The paper (1), apart from describing the new Metaxa version, also brings a very thorough evaluation of the software, compared to other tools for taxonomic classification implemented in QIIME (2). In short, we show that:

  • Metaxa2 can make trustworthy taxonomic classifications even with reads as short as 100 bp
  • Generally, the performance is reliable across the entire SSU rRNA gene, regardless of which V-region a read is derived from
  • Metaxa2 can reliably recapture species composition from short-read metagenomic data, comparable with results of amplicon sequencing
  • Metaxa2 outperforms other popular tools such as Mothur (3), the RDP Classifier (4), Rtax (5) and the QIIME implementation of Uclust (6) in terms of proportion of correctly classified reads from metagenomic data
  • The false positive rate of Metaxa2 is very close to zero; far superior to many of the above mentioned tools, many of which assume that reads must derive from the rRNA gene

Metaxa2 can be downloaded here. We have already used it for around two years internally, and it forms the base of the taxonomic classifications in e.g. our recently published paper on antibiotic resistance in a polluted Indian lake (7).

References

  1. Bengtsson-Palme J, Hartmann M, Eriksson KM, Pal C, Thorell K, Larsson DGJ, Nilsson RH: Metaxa2: Improved Identification and Taxonomic Classification of Small and Large Subunit rRNA in Metagenomic Data. Molecular Ecology Resources (2015). doi: 10.1111/1755-0998.12399 [Paper link]
  2. Caporaso JG, Kuczynski J, Stombaugh J et al.: QIIME allows analysis of high-throughput community sequencing data. Nature Methods, 7, 335–336 (2010).
  3. Schloss PD, Westcott SL, Ryabin T et al.: Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Applied and Environmental Microbiology, 75, 7537–7541 (2009).
  4. Wang Q, Garrity GM, Tiedje JM, Cole JR: Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Applied and Environmental Microbiology, 73, 5261–5267 (2007).
  5. Soergel DAW, Dey N, Knight R, Brenner SE: Selection of primers for optimal taxonomic classification of environmental 16S rRNA gene sequences. The ISME Journal, 6, 1440–1444 (2012).
  6. Edgar RC: Search and clustering orders of magnitude faster than BLAST. Bioinformatics, 26, 2460–2461 (2010).
  7. Bengtsson-Palme J, Boulund F, Fick J, Kristiansson E, Larsson DGJ: Shotgun metagenomics reveals a wide array of antibiotic resistance genes and mobile elements in a polluted lake in India. Frontiers in Microbiology, 5, 648 (2014).

My colleague Henrik Nilsson has been interviewed by the ResearchGate news team about the recent effort to better annotate ITS data for plant pathogenic fungi. It’s an interesting read, and I think Henrik nicely underscores why large-scale efforts for improving and correcting sequence annotations are important. You can read the interview here, and the paper they talk about is referenced below.

Nilsson RH, Hyde KD, Pawlowska J, Ryberg M, Tedersoo L, Aas AB, Alias SA, Alves A, Anderson CL, Antonelli A, Arnold AE, Bahnmann B, Bahram M, Bengtsson-Palme J, Berlin A, Branco S, Chomnunti P, Dissanayake A, Drenkhan R, Friberg H, Frøslev TG, Halwachs B, Hartmann M, Henricot B, Jayawardena R, Jumpponen A, Kauserud H, Koskela S, Kulik T, Liimatainen K, Lindahl B, Lindner D, Liu J-K, Maharachchikumbura S, Manamgoda D, Martinsson S, Neves MA, Niskanen T, Nylinder S, Pereira OL, Pinho DB, Porter TM, Queloz V, Riit T, Sanchez-García M, de Sousa F, Stefaczyk E, Tadych M, Takamatsu S, Tian Q, Udayanga D, Unterseher M, Wang Z, Wikee S, Yan J, Larsson E, Larsson K-H, Kõljalg U, Abarenkov K: Improving ITS sequence data for identification of plant pathogenic fungi. Fungal Diversity, Volume 67, Issue 1 (2014), 11–19. doi: 10.1007/s13225-014-0291-8 [Paper link]

Another paper I have co-authored related to the UNITE database for fungal rDNA ITS sequences is now published as an Online Early article in Fungal Diversity. The paper describes an effort to improve the annotation of ITS sequences from fungal plant pathogens. Why is this important? Well, apart from fungal plant pathogens being responsible for great economic losses in agriculture, the paper is also conceptually important as it shows that together we can accomplish a substantial improvement to the metadata in sequence databases. In this work we have hunted down high-quality reference sequences for various plant pathogenic fungi, and re-annotated incorrectly or insufficiently annotated ITS sequences from the same fungal lineages. In total, the 59 authors have made 31,954 changes to UNITE database data, on average 540 changes per author. While one, or a few, persons could not feasibly have made this effort alone, this work shows that in larger consortia vast improvements can be made to the quality of databases, by distributing the work among many scientists. In many ways, this relates to proposals to “wikify” GenBank, and after Rfam and Pfam it might now be time to take the user-contribution model to, at least, the RefSeq portion of GenBank, which despite its description as being “comprehensive, integrated, non-redundant, [and] well-annotated” still contains errors and examples of non-usable annotation. More on that at a later point…

Paper reference:

Nilsson RH, Hyde KD, Pawlowska J, Ryberg M, Tedersoo L, Aas AB, Alias SA, Alves A, Anderson CL, Antonelli A, Arnold AE, Bahnmann B, Bahram M, Bengtsson-Palme J, Berlin A, Branco S, Chomnunti P, Dissanayake A, Drenkhan R, Friberg H, Frøslev TG, Halwachs B, Hartmann M, Henricot B, Jayawardena R, Jumpponen A, Kauserud H, Koskela S, Kulik T, Liimatainen K, Lindahl B, Lindner D, Liu J-K, Maharachchikumbura S, Manamgoda D, Martinsson S, Neves MA, Niskanen T, Nylinder S, Pereira OL, Pinho DB, Porter TM, Queloz V, Riit T, Sanchez-García M, de Sousa F, Stefaczyk E, Tadych M, Takamatsu S, Tian Q, Udayanga D, Unterseher M, Wang Z, Wikee S, Yan J, Larsson E, Larsson K-H, Kõljalg U, Abarenkov K: Improving ITS sequence data for identification of plant pathogenic fungi. Fungal Diversity Online early (2014). doi: 10.1007/s13225-014-0291-8 [Paper link]

Our paper on the most recent developments of the UNITE database for fungal rDNA ITS sequences has just been published as an Early view article in Molecular Ecology. In this paper, we aim to ease two of the major problems facing the identification of newly generated fungal ITS sequences: the lack of a sufficiently goof reference dataset, and the lack of a way to refer to fungal species without a Latin name. As part of a solution, we have introduced the term species hypothesis for all fungal species represented by at least two ITS sequences. The UNITE database has an easy-to-use web-based sequence management system, and we encourage everybody that can improve on the annotations or metadata of a fungal lineage to do so.

My main contribution on this paper has been to tailor ITSx functionality for the UNITE database, so that ITS data could be more easily processed for the Species Hypotheses.

Paper reference:
Kõljalg U, Nilsson RH, Abarenkov K, Tedersoo L, Taylor AFS, Bahram M, Bates ST, Bruns TT, Bengtsson-Palme J, Callaghan TM, Douglas B, Drenkhan T, Eberhardt U, Dueñas M, Grebenc T, Griffith GW, Hartmann M, Kirk PM, Kohout P, Larsson E, Lindahl BD, Lücking R, Martín MP, Matheny PB, Nguyen NH, Niskanen T, Oja J, Peay KG, Peintner U, Peterson M, Põldmaa K, Saag L, Saar I, Schüßler A, Senés C, Smith ME, Suija A, Taylor DE, Telleria MT, Weiß M, Larsson KH: Towards a unified paradigm for sequence-based identification of Fungi. Accepted in Molecular Ecology. doi: 10.1111/mec.12481 [Paper link]

One potential use for Metaxa (paper) is to include it in a pipeline for classification of SSU rRNA in metagenomic data (or other environmental sequencing sets). However, as Metaxa is provided from this site, it only classifies SSUs to the domain level (archaea, bacteria and eukaryotes, with the addition of chloroplasts and mitochondria). It is also able to do some (pretty rough) species guesses using the “--guess_species T” option. An easy solution to implement would be to pass the Metaxa output, e.g. “metaxa_output.bacteria.fasta” to BLAST, and compare all these sequences to the sequences in e.g. the SILVA or GreenGenes database. There is, however, a way to improve this, which uses Metaxa’s ability to compares sequences to custom databases. In this tutorial, I will show you how to achieve this.

Before we start, you will of course need to download and install Metaxa, and its required software packages (BLAST, HMMER, MAFFT). When you have done this, we can get going with the database customization. I will in this tutorial use the SILVA database for SSU classification. However, the basic idea for the tutorial should be easily applicable to GreenGenes and other rRNA databases as well.

  1. Visit SILVA through this link, and download the file named “SSURef_106_tax_silva.fasta.tgz”. The file is pretty big so it may take a while to download it. If you’re running Metaxa on a server, you’ll have to get the SILVA-file to the server somehow.
  2. Unzip and untar the file (Mac OS X makes this neatly by doubleclicking the file, on linux you can do it on the command line by typing “tar -xvzf SSURef_106_tax_silva.fasta.tgz“). This will give you a FASTA-file.
  3. The FASTA-file needs to be prepared a bit for Metaxa usage. First, we need to give Metaxa identifiers it can understand. Metaxa identifies sequences’ origins by the last character in their identifier, e.g. “>A16379.1.1496.B”. Here, “.B” indicates that this is a bacterial sequence. We are now going to use the unix command sed to process the file and insert the appropriate identifiers.
    1. We begin with the archaeal sequences. To get those straight, we type:
      sed "s/ Archaea;/.A - Archaea;/" SSURef_106_tax_silva.fasta > temp1
      Notice that we direct the output to a temporary file. It is bad practice to replace the input file with the output file, so we work with two temp-files instead.
    2. The next step is also easy, now we find all eukaryote sequences and add E:s to the identifiers:
      sed "s/ Eukaryota;/.E - Eukaryota;/" temp1 > temp2
    3. Now it becomes a little more complicated, as SILVA classes mitochondrial and chloroplast SSU sequences as subclasses of bacteria. However, there is a neat little trick we can use. First we do the same with the bacterial sequences as with the archaeal and eukaryote:
      sed "s/ Bacteria;/.B - Bacteria;/" temp2 > temp1
    4. Now, we can use two a little more complicated commands to annotate the mitochondrial and chloroplast sequences:
      sed "s/\.B - \(Bacteria;.*;[Mm]itochondria;\)/.M - \1/" temp1 > temp2
      sed "s/\.B - \(Bacteria;.*;[Cc]hloroplast;\)/.C - \1/" temp2 > temp1
    5. We also need to get “rid” of the unclassified sequences, by assigning them to the “other” origin (O):
      sed "s/ Unclassified;/.O - Unclassified;/" temp1 > temp2
  4. That wasn’t too complicated, was it? We can now check the number of different sequences in the file by typing the pretty complicated command:
    grep ">" temp2 | cut -f 1 -d " " | rev | cut -f 1 -d "." | sort | uniq -c
    If you have been working with the same files as me, you should now see the following numbers:
    23172 A
    471949 B
    3712 C
    55937 E
    534 M
    226 O
  5. At this stage, we need to remove the full taxonomy from the FASTA headers, as Metaxa cannot handle species names of this length. We do this by typing:
    sed "s/ - .*;/ - /" temp2 > temp1
  6. We can now change the temp-file into a FASTA file, and delete the other temp-file:
    mv temp1 SSURef.fasta
    rm temp2
  7. We now need to configure Metaxa to use the database. First, we format a BLAST-database from the FASTA-file we just created:
    formatdb -i SSURef.fasta -t "SSURef Metaxa DB" -o T -p F
  8. With that done, we can now run Metaxa using this database instead of the classification database that comes with the program. By specifying that we want to guess the species origin of sequences, we can get (as accurate as SILVA lets us be) which species each sequence in our set come from. We do this by using the -d and the --guess_species options:
    metaxa -i test.fasta -d SSURef.fasta -o TEST --guess_species T --cpu 2
    The input in this case was the test file that comes with Metaxa. Note also that we’re using two CPUs to get multithreaded speeds. Remember that you must provide the full (or relative) path to the database files we just created, if you are not running Metaxa from the same directory as the database resides in.
  9. The output should now look like this (taken from the bacterial file):
    >coryGlut_Bielefeld_dna Bacterial 16S SSU rRNA, best species guess: Corynebacterium glutamicum
    CGAACGCTG...
    >gi|116668568:792344-793860 Bacterial 16S SSU rRNA, best species guess: Arthrobacter sp. J3.40
    TGAACGCTG...
    >gi|117927211:c1399163-1397655 Bacterial 16S SSU rRNA, best species guess: Acidothermus cellulolyticus
    >CGAACGCTG...

    And so on. As you can see the species names are now located at the end of each definition line, and can easily be extracted using sed, e.g. “grep ">" TEST.bacteria.fasta | sed "s/.*: //"“.

And that’s it. It’s pretty simple, and can easily be scripted. In fact, I have already made the bash script for you. That means that the short version is, download the script, download the sequence file from SILVA, move into the directory you have downloaded the file to and run the script by typing: ./prepare_silva_for_metaxa.sh

A few notes at the end. The benefit of using this approach is that we maintain the sorting capabilities, marking of uncertain sequences and error checking of Metaxa, but we don’t have to add another BLAST step after Metaxa has finished. However, as this database we create is a lot bigger than the database that comes with Metaxa, the running time of the classification step will be substantially longer. This is in most cases acceptable, as that time is the same as the time it would have taken to run BLAST on the Metaxa output. It should also be noted that this approach limits Metaxa’s ability of classifying 12S sequences, as there are no such sequences in SILVA. Good luck with classifying your metagenome SSUs (and if you use Metaxa in your research, remember to cite the paper)!