Metaxa2 User's Guide - Microbiology, Metagenomics and
Contents
1. In this way all sequences get a taxonomic affiliation at a trustworthy taxonomic level This data is written to a file with the suffix taxonomy txt Metaxa2 s default settings should be usable in most situations However since the software has been tweaked to accommodate for analysis of really large datasets you should consider if they are suitable for your purposes and for your data set If the data set is small this can be done by running the software multiple times on the data with different settings and analyse the outcome On larger data sets it might be more feasible to only run Metaxa2 on a subset of the sequences for testing The graphical output is very useful for determining whether Metaxa2 performs as desired on long read gt 400 bp data as the positions of the found conserved domains can be easily investigated If domains are missing the criteria might be set to be too stringent If they are not in sequential order from V11 to V9r that might be an indication that there is something wrong with the input sequences The HMMER program hmmsearch used by Metaxa2 uses heuristic filters to increase the search speed Metaxa2 runs hmmsearch as it is with the max option disabled To turn off all heuristic filters of HMMER Metaxa2 can be run with the heuristics F option This increases detection power at the cost of speed 7 Running Metaxa2 s analysis steps separately Metaxa2 s analysis procedure is d2. This custom set can be any type of profiles but the profiles must be named according to the convention in the other HMM files beginning with V1I and ending with V9r taxonomy complete Itis possible to add the underlying taxonomic data which the taxonomic predictions are based on by using the taxonomy complete option instead of setting it to taxonomy T This data is added to each entry in the taxonomy txt output file 9 License information This program is free software you can redistribute it and or modify it under the terms of the GNU General Public License as published by the Free Software Foundation either version 3 of the License or at your option any later version This program is distributed in the hope that it will be useful but WITHOUT ANY WARRANTY without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License for more details You should have received a copy of the GNU General Public License along with this program in a file called license txt If not see http www gnu org licenses Copyright C 2011 2013 Johan Bengtsson Palme et al
3. and would be filtered out On T by default Specifies the distance between the sequence pairs often referred to as the insert size Changing this value has little impact on Metaxa2 s actual performance Default 150 Sequence selection options t b bacteria a archaea e eukaryota m mitochondrial c chloroplast A all E value S value N value M value R value H value selection_priority score sum domains eval search_eval value search_score value blast_eval value blast_score value blast_wordsize value Set of profiles to use for the search comma separated Accepts any list of sets e g bacteria chloroplast m c or eukaryota Can be used to restrict the search to only a few SSU LSU types to save time if one or more of the origins are not relevant to the dataset under study Default is to use all the all option Domain E value cutoff a sequence must obtain in the HMMER based step to be included in the output Default 1 Domain score cutoff for a sequence must obtain in the HVMER based step to be included in the output Default 12 The minimum number of domains that must match a sequence for it to be included in the output Setting the value lower than two will increase the number of false positives while increasing it above two will decrease Metaxa s detection abilities on fragmentary data Default 2 Num
4. i file Nucleotide FASTA FASTQ input file to investigate Metaxa2 accepts both aligned and unaligned FASTA If no input is specified Metaxa2 will read sequences from standard input which means that FASTA sequences can be piped into Metaxa2 o file 1 file 2 file pairfile file f a auto f fasta q fastq p paired end pa paired fasta g ssu lsu p directory d database date T or F plus T or F Base for the file names of the output files Suffixes will be added automatically Defaults to metaxa_out DNA FASTA FASTQ input file containing the first reads in the read pairs to investigate Use instead of the i option DNA FASTA FASTQ input file containing the second reads in the read pairs to investigate Use instead of the i option and only together with the 1 option Note that the reads in the files must be in the same order As an alternative to using the 1 and 2 options the user might use i and pairfile instead which has the same meaning Specifies the format of the input file s By default Metaxa2 will try to auto detect the format auto Synonymous to format Specifies if Metaxa2 should identify and extract SSU or LSU rRNA genes By default Metaxa2 will look for SSU genes ssu A path to a directory containing HMM profile collections representing SSU rRNA conserved regions By default Metaxa assumes to find the databases in the metaxa_db directory l
5. into the binaries directory and move all of its contained files into your preferred bin directory usually either usr local bin or your own bin directory home user bin The HMMER package should now be installed on your computer you can check this by typing hmmscan h in the terminal and press enter you should now see HMMER output Download and install the BLAST package ftp ftp ncbi nlm nih gov blast executables release LATEST for sequence similarity searches Metaxa2 works with both the legacy version of BLAST as well as BLAST It should work with any version of BLAST starting with version 2 2 and better Download the BLAST package for your operating system to your preferred directory Open a command line terminal move into the directory with cd home user and unpack the tarball with tar xvfz blast 2 2 24 platform tar gz Move into the bin directory inside the newly created BLAST directory and move all of its contained files into your preferred bin directory Alternatively you can add the BLAST bin directory to your PATH The BLAST package should now be installed on your computer you can check this by typing blastall in the terminal and press enter you should now see the listing of BLAST options Download and install the MAFFT http mafft cbrc jp alignment software for multiple alignment Metaxa2 as previous versions of Metaxa relies on MAFFT version 6 MAFFT is not critical for Metaxa2 s core fun
6. sequences of uncertain origin to a directory with the suffix _alignments The user may specify to instead align all SSU sequences by using the align all option note that this would increase the runtime significantly The five best BLAST matches are aligned to the query sequence and saved to an aligned FASTA file with the name lt query identifier gt aligned fasta Chimeric sequences If the option allow_reorder is turned off Metaxa2 will save an additional FASTA file containing sequences that are suspected to be chimeric These are sequences with domains located in the wrong order This is useful on full length or near full length data sets but should not be used on short reads as it could increase the number of false negatives when run on short sequences Raw data If the option to save all raw data is turned on Metaxa2 will save all data from the pre processing HMMER search BLAST search as well as a file of raw statistics into a directory with the suffix _metaxa_raw_output 5 Metaxa2 Taxonomic Traversal Tool In addition to the improved classifier Metaxa2 also introduces a new bundled tool to further analyse the taxonomy output This tool called the Metaxa2 Taxonomic Traversal Tool metaxa2_ttt summarizes the taxonomic output of Metaxa at different taxonomic levels Simply put the traversal tool goes through the taxonomic predictions in the taxonomy txt output file and counts the number o
7. somewhat This should be kept in mind when using Metaxa2 on longer reads which might require tweaking of the default options one may for example compare to the default options of Metaxa 1 1 2 Quality pre filtering of reads in FASTQ format Metaxa can use the quality information in the FASTQ files to filter out low quality reads or read pairs This is controlled by a range of options described in part three of the manual Usage and Commands Support for the modern BLAST package Metaxa now incorporates the long sought after support for the NCBI BLAST package enabling users to move on to what 1s deemed the future of BLAST Compatibility with HMMER 3 1 Not only does Metaxa2 bring BLAST compatibility it also secures support for the upcoming version of HMMER released in beta at the time of writing Changed default priority for scoring best HMMER match As part of the optimization for larger data sets and shorter reads Metaxa s priority system for determining the best HMMER match has now been changed from sum to score To switch back to Metaxa 1 behaviour one can use the selection_priority sum option Be aware of that this might influence other parameters as well HMMER s heuristics are now used by default Another optimization for larger data sets is that the heuristics of HMMER has now been turned on by default To switch the heuristics of use the heuristics F option Note however that this
8. will make HMMER much slower on large data sets 3 Usage and commands For the very impatient only follow the brief installation instructions in the file README txt To check for SSU rRNA sequences in the file test fasta you would then type metaxa i test fasta o test on the command line To check for LSU rRNA sequences type metaxa i test fasta o test g lsu instead For all other users Metaxa2 accepts input in the FASTA and FASTQ formats As it pre processes the input sequences it is possible to input both aligned and unaligned FASTA files containing both DNA and RNA sequences By default Metaxa2 outputs ten files one summary file of the entire run one graphical representation of hits one FASTA file of all identified SSU sequences one FASTA file for each of the six possible origins and one file containing the predicted taxonomic origin of the extracted sequences To list all the available options for Metaxa2 type metaxa2 help You can use the test fasta file that comes bundled with the software for a test run This file contains 50 randomly selected SSU entries ten of each origin 50 randomly selected LSU entries and 10 non SSU non LSU sequences In the simplest case Metaxa2 is run by metaxa2 i input_file o output Below is a listing of all options Metaxa2 accepts Boolean options can be turned on with T true or 1 and off using F false or 0 Main options
9. PATH by adding the line export PATH PATH HOME bin to the file profile in your home directory The process of adding items to one s PATH varies among systems and shells Close the terminal and open a new one for this change to take effect Perl needs to be installed on the computer Most Unix based systems including Linux and MacOS X have Perl pre installed You can check this by opening a command line terminal and type perl v In case Perl is not installed you have to download http www perl org and compile the program Download and install HMMER version 3 http hmmer janelia org software Version 2 of Metaxa relies on HMMER version 3 just as the previous Metaxa versions Metaxa will not work with earlier versions of HMMER although it will work with the HMMER 3 1 beta Download the HMMER package source code to your preferred directory such as home user Open a command line terminal move into the directory with cd home user and unpack the tarball with tar xvfz hmmer 3 0 tar gz Now you need to compile HMMER from source files To compile it from source enter the new directory and follow the installation instructions in the file INSTALL If you have trouble compiling HMMER you can try to use the pre compiled binaries available at the HMMER home page After download and unpacking of the tarball the binaries are located in the binaries directory contained within the newly created HMMER directory Move
10. User s guide Manual for Metaxa2 This is a guide to install and use the software utility Metaxa2 The software is written for Unix like platforms and should work on nearly all Linux based systems as well as MacOS X Contents of this manual p Detailed installation instructions Changes from previous versions of Metaxa Usage and commands Output files Metaxa Taxonomic Traversal Tool Internal changes in Metaxa2 Running Metaxa2 s analysis steps separately Undocumented features CO 02 gt Us O OW AR Sh License information 1 Detailed installation instructions The README txt file bundled with the script provides a quick installation guide In order to install certain packages you might need to have superuser privileges For installation on Mac you will have to install the Apple Xcode package available on your MacOS X System DVD in order to be able to compile programs Please talk to your system administrator if you feel unsure about these steps Note that the packages are mandatory and that you should not proceed unless these criteria are fulfilled If you don t have superuser privileges on your machine Create a directory within your user directory e g home user bin and to store all required binaries there By adding this directory to your PATH any software placed in the directory will behave as if installed for all users using superuser privileges If you use the bash shell you can add a bin directory to your
11. an be accessed by running Metaxa with the save_raw T option The metaxa2_x program reads the concatenated FASTA file and the pair info to create an output file containing the identified rRNA sequences In this file it also inserts an insert sequence consisting of repeated N characters This file is then used as a regular FASTA file as input to metaxa2_c Thus metaxa2_c really does not consider the paired reads it looks at each concatenated entry as one single long FASTA sequence with a stretch of unknown nucleotides in it The Metaxa2 Extractor metaxa2_x has two major changes to it since the last version First of all it is now gene agnostic that is it is not hardcoded for SSU genes anymore In theory that means that any marker gene or region could now be used with Metaxa2 in practice however it only means added support for the LSU rRNA gene This can be seen in the naming of the HMM profiles which now are ordered sequentially without notions of e g V1 and V3r Secondly the extractor has a different way of handling the input sequences partially inspired by the work we did on ITSx http microbiology se software itsx In principle this means that Metaxa2 does not process sequences sequentially Instead batches of sequences the number dependent on the available system memory are kept in RAM and processed separately If you choose the to use the save_raw T option you may notice that there are fi
12. ber of top BLAST matches that should be considered in classification This setting also affects the number of matches used for taxonomic classification Default 5 Reliability score cutoff for the taxonomic classification Entries not satisfying the cutoff will be classified at the level above in the taxonomic hierarchy until the reliability score is above the threshold Default 80 The number of points that the predicted origin of the Metaxa Extractor is given Default is the same as the number of sequences used for classification M option above which is set to 5 by default Determines what will be of highest priority when assessing the origin of the sequence Options are score which uses the average score of the found hits sum which sums the scores for each profile match and divides the sum by the number of profiles of the given type domains which uses the number of domains retrieved of a given type eval which uses the average E value of the found hits Default is to use score The actual E value cutoff used in the HMMER search High numbers may slow down the process Should never be set to a lower value than the E option Cannot be used in combination with the search_score option Default is to use score cutoff see search_score below not E value The score cutoff used in the HMMER search Low numbers may slow down the process Should never be set to a higher number than the S option Cannot be used in com
13. bination with the search_eval option Default 0 The E value cutoff used in the BLAST search High numbers may slow down the process Cannot be used in combination with the blast_score option Default is 1e 15 The score cutoff used in the BLAST search Low numbers may slow down the process Cannot be used in combination with the blast_eval option Default is to use E value cutoff see blast_eval above not score The word size used for the BLAST based classification Lower numbers will slow down the process significantly while higher numbers may allow_single_domain e value score or F allow_reorder T or F complement T or F cpu value multi_thread T or F heuristics T or F megablast T or F Output options summary T or F graphical T or F fasta T or F table T or F taxonomy T or F taxlevel integer not_found T or F align a all u uncertain n none truncate T or F guess_species T or F potentially decrease classification accuracy Default is 14 Allow inclusion of sequences that only find a single domain given that they meet the more stringent E value and score thresholds specified By default single domains are allowed with E value cutoff 1e 10 and score cutoff 0 1e 10 0 Allows profiles not to be in the expected order 1 9 on the extracted sequences If turned off a file of potential chimeric sequences w
14. by preliminary origin Archaea 0 Bacteria 0 Eukaryota 0 Chloroplast 100 Mitochondria 0 Other 0 Number of SSU rRNA sequences to be classified by Metaxa 100 Number of SSU rRNA having at least one database match 100 Number of SSU rRNA successfully classified by Metaxa 100 Number of uncertain classifications of SSU rRNA sequences 0 Total number of classifications made by Metaxa 100 Number of SSU rRNA sequences assigned to each origin Archaea 0 Bacteria 0 Eukaryota 0 Chloroplast 100 Mitochondria 0 Uncertain 0 Sequences of chloroplast origin 16S Acorus_americanus_AcamCr001 Aethionema_cordifolium_AecoCr001 Welwitschia mirabilis WemicC_r001 Zea_mays_ZemaCr113 Metaxa run finished at Tue Jul 14 10 08 42 2013 Taxonomy table One of the new features of Metaxa2 is the much improved ability to make taxonomic predictions of identified rRNA sequences The results of these predictions are written to a file with the suffix taxonomy txt Each input sequence is represented by one line in this tab separated file with five columns Column Description ID The identifier of the query sequence Classification The taxonomic tree for which Metaxa2 has been able to a reliable prediction Identity The per cent identity to the best BLAST match in the database Length The length of the alignment between the input sequence and the best BLAST match Reliability score The number of conserved domains for the most like
15. chondrial c chloroplast A all o other r value Reliability cutoff Entries with reliability scores below this cutoff will be classified as unknown Default 0 value Length cutoff in bp for the best BLAST hit Entries below this cutoff will be classified as unknown Default 50 d value Per cent identity cutoff for the best BLAST hit Entries below this cutoff will be classified as unknown Default 0 m integer Maximum resolution level for taxonomic traversal Providing a zero 0 to this option will remove this limit Default is to have no upper limit for traversal 0 n integer Minimum resolution level for taxonomic traversal The kingdom level is defined as 1 and is the lowest possible setting Default 1 s T or F If true metaxa2_ttt will investigate only the last taxonomic level which in the best case will represent a species resolution Note that this setting will classify sequences at very different taxonomic levels When true metaxa2_ttt will only output the level_1 txt output file Default is off F remove_na T or F If true sequence entries with no BLAST hits will be set to Unknown Default is on T Output options summary T or F Outputs a summary of results On T by default lists T or F Outputs a lists of counts for different taxa one file for each traversal level On T by default separate T or F Outputs the taxonomy file used as i
16. ctions but is used for automatically creating alignments of uncertain sequences Instructions for installing MAFFT are available on the MAFFT download page Go to http microbiology se software metaxa 2 in order to download the Metaxa 2 package Download it to your preferred directory Unpack the downloaded tarball with tar xvfz metaxa2_2 0 tar gz A directory called Metaxa2 will be created You will see the following files and directories inside it metaxa2 metaxa2_x metaxa2_c metaxa2_ttt install_metaxa2 the metaxa2_db directory containing the Hidden Markov Models and a BLAST database the user s guide the README txt file the license txt file as well as test input files Enter the directory and type install_metaxa2 Press enter and follow the on screen instructions You will be prompted for whether you have superuser privileges and where you want Metaxa to be installed If Metaxa2 is successfully installed you should see its help message when typing the command metaxa2 help 2 Changes from previous versions of Metaxa Metaxa introduces a couple of new features over Metaxa 1 1 2 as well as changing the default behaviour of some options The main changes will be outlined below Extraction and classification of LSU rRNA sequences One of the major new features in Metaxa2 is the addition of a second barcoding gene in addition to the SSU gene the large subunit rRNA LSU also known as 23S rRNA in prokaryotes a
17. e definition line in the example above An uncertain sequence could look like this gt AABL01000014 4508 5931 Putative Chloroplast 16S SSU rRNA GAACGCTAGAAATATACATTACACATGCAAATTTATGATAATATCATAGTGAATAGGTGA The extraction file contains all sequences identified as rRNA by metaxa2_x the first step of the analysis The sequence entries in that file contain information on what domains that were found and what origin that is most likely base on the profile search An example is shown below gt A16379 1 1496 B Predicted Bacterial 16S SSU rRNA 1447 bp From domain V1l to V9r on main strand Found domains V11 V21 V2r V31 V3r V4l V4r V51 V5r V6l V71 V81 V8r V91 V9Or CAGGCTTAACACATGCAAGTCGAACGGTAGCACGAAGGACTTGCTCCTTGGGTGACGAGT Summary A summary of the Metaxa2 run is written to a file with the suffix summary txt In this file the statistics of the run is collected as are the starting and ending times for the run Also lists of the identifiers of extracted SSU sequences are written to this file one list for each origin The first section of the file shows the data from the extraction step The second section is associated with the second classification step After the second section the lists of entries of different origins are found An example of parts of a summary file is shown below Number of sequences in input file 100 Sequences detected as SSU rRNA by Metaxa 100 On main strand 91 On complementary strand 9 SSU sequences
18. entry is associated with a lineage and three numbers the per cent identity to the best matching BLAST hit the length of that alignment and a reliability score The reliability score is calculated based on the per cent identity to the best BLAST hit and how divergent the rest of the BLAST hits are from the first one The maximum value of this score is 100 This score is only given if all five BLAST matches are 100 identical to the input sequence and all those matches represent the same taxonomic lineage This means that this situation rarely if ever occurs on real data Instead scores above 80 should be considered trustworthy The reliability score can then be used to filter out uncertain entries when summarizing the taxonomic predictions with the Metaxa2 Taxonomic Traversal Tool metaxa2_ttt described below The Metaxa2 Taxonomic Traversal Tool metaxa2_ttt Because of the extended abilities of Metaxa2 to classify SSU and LSU sequences more in detail a new tool to investigate the organismal content of the sample at different taxonomic levels has been included in the Metaxa2 package This tool is called the Metaxa2 Taxonomic Traversal Tool since it does exactly that it traverses the taxonomy txt output file and reports Metaxa s taxonomic predictions according to specified cut offs The complete usage of the metaxa2_ttt tool is described in part five of this manual Metaxa2 Taxonomic Traversal Tool In short the traversal
19. esentations are explained in the table below Feature Description Part of the sequence without any conserved domain variable region Vil Start of a conserved domain Continuation of a conserved domain gt Indicates that one conserved domain goes into the next Domains are normally not overlapping so this could be an indication of a bad input sequence The line of asterisks indicates the end of one set of matches Note that the graph should be viewed with a non proportional font such as Courier if loaded into e g Word Extraction results table The full results of the Metaxa2 extraction is saved to a file with the suffix extraction results This file consists of tab separated columns containing various information on each SSU sequence found The file can be easily imported into programs such as Excel The contents of the columns from left to right are explained in this table Column ID Length Origin Strand Domains Average E value Average score Start End First domain Last domain Chimera Specific origin information Description The identifier of the query sequence The length of the query sequence A one letter abbreviation of the sequence origin A archaeal B bacterial C chloroplast E eukaryote M mitochondrial 16S N mitochondrial 12S A zero 0 if the SSU LSU was found on the main strand a one 1 if it was found on the complementary strand The number of conserved do
20. f entries associated with each taxonomic level The levels are roughly corresponding to kingdoms phyla classes orders families genera species and subspecies in some cases followed by more specific annotations The traversal tool can also filter the output according to reliability score alignment length per cent identity to the best BLAST match and or taxonomic group The output of metaxa2_ttt consists of a number of tab separated text files containing group counts at different taxonomic levels by numbers and a summary file with the suffix taxonomy summary txt Each of the count file have the following format Bacteria Actinobacteria Actinobacteria 5 Bacteria Firmicutes Bacilli 14 Bacteria Firmicutes Clostridia 1 Bacteria Proteobacteria Betaproteobacteria 2 Bacteria Unclassified Bacteria 3 Each line here represents one node in the taxonomic tree and the second column contains the number of entries associated with that node The options for metaxa2_ttt are given in this table Options i file Metaxa2 taxonomy output file to process with suffix taxonomy txt to use as input for metaxa2_ttt o file Base for the file names of the output files Suffixes will be added automatically Defaults to metaxa_ttt_out t b bacteria a Include only classifications of this taxonomic type s Several types can be archaea e specified separated by comma Default is to use all types all eukaryota m mito
21. ith profile matches in the wrong order is written allowing for rudimentary chimera detection This can and should be used on full length sequences On fragmented sequences however the risk of missing true positives increases if this option is turned off On T by default If on Metaxa2 checks both DNA strands for matches to HMM profiles On T by default The number of CPU threads to use Metaxa performs significantly faster using more CPUs Default is 1 Multi thread the HMMER search On T by default if the number of CPUs is larger than one cpu option gt 1 else off F Selects whether to use HMMER s heuristic filtering On T by default Turning this setting off will decrease speed but increase precision Uses megablast for classification for better speed but less accuracy Off F by default If on Metaxa outputs a summary of results File suffix is ssummary txt On T by default If on Metaxa outputs graphical text representations of where in each sequence the conserved domains were found File suffix is graph On T by default If on FASTA formatted files containing the extracted SSU sequences are written One file for each origin is written plus an extraction file containing all SSUs identified in the first analysis step On T by default If on Metaxa saves table format output of results separately for HAMER and BLAST output Note that neither of these outputs is the actual outpu
22. ivided into three steps pre processing extraction and classification These steps are normally run in sequence by running the metaxa2 command However they can also be run separately if the user wishes To run the extraction step independently use the metaxa2_x command This command takes a subset of the Metaxa2 options other options will be ignored To see the available options for the metaxa2_x command type metaxa2_x help on the command line To run the classification step on a set of known rRNA sequences use the command metaxa2_c The options for metaxa2_c can be seen by typing metaxa2_c help on the command line Note that the output files obtained when running each step separately will be slightly different than obtained through running the entire Metaxa2 pipeline Additionally FASTQ and paired end file processing is not supported when running the analysis steps separately 8 Undocumented features Metaxa has three undocumented options that can be activated but they are considered experimental and should be used with caution One allows you to use pre calculated hmmscan results and feed into Metaxa2 another allows using a set of additional HMM profiles for the SSU extraction and the final allows seeing the underlying taxonomic data justifying the taxonomic predictions Undocumented options t o other It is possible to supply an additional set of HMM profiles in an O hmm file within the HMMs directory
23. les containing 1 and higher numbers among the raw data files Those are the batches processed by Metaxa2 This way of handling sequences greatly speed up Metaxa2 s post processing of HMMER output which becomes a time limiting step in typical Illumina datasets Minor such changes were introduced already in Metaxa 1 1 but have now been fully implemented to accommodate for increasingly bigger datasets The sequence handling has also gone through some minor changes to be able to deal with paired end sequences However this should not affect the behaviour on single end sequences The classifier metaxa2_c has undergone a complete overhaul when it comes to make taxonomic predictions The new classification system is built upon the taxonomic information of the by default five best BLAST matches to each rRNA in the input data this can be changed using the M option For each rRNA entry Metaxa2 compares the taxonomic affiliation of the top BLAST match with the second one and so on In each comparison the per cent identity to the query is taken into account If the BLAST matches point to the same taxonomic origin the query sequence gets a taxonomic affiliation with a high reliability score close to 100 If not that is if the score by default is below 80 specified by the R option the comparison is repeated at the taxonomic level above e g genus if the last comparison was made on species level until the score is above 80
24. ll options bugs Displays the bug fixes and known bugs in this version of Metaxa license Displays licensing information 4 Output files Metaxa outputs a number of files depending on what is selected by the user see Usage and Commands above By default seven FASTA files a table of taxonomic classifications a file containing graphical representation of putative SSU LSU sequences and a summary file is written In addition tables of BLAST and HMMER results lists of non targeted entries and sequence alignments can be written on request by the user There is also an option to preserve all the intermediate data generated by the HMMER and BLAST searches FASTA output Metaxa2 generates one FASTA file for each origin archaea bacteria eukaryota chloroplast and mitochondria one file containing sequences of uncertain origin and one file with all rRNA sequences identified and extracted in the first step Sequences in these files are marked according to their origin Sequences whose origin Metaxa could not establish with certainty but for which enough data were available to allow a qualified guess as to the origin of the sequences are marked with a character at the end of the definition line A certain sequence may look like this gt gi 117927211 Bacterial 16S SSU rRNA GTTTGATCCTGGCTCAGGACGAACGCTGGCGGCGTGCTTAACACATGCAAGTCGAGCGGA Note that Metaxa2 has added the type of the SSU sequence Bacterial 16S SSU rRNA to th
25. lt name of matching profile gt lt score gt lt E value gt As in the graphical output file the table file is divided into sections Each section represents one group of sequences and begins with the line X matches on main strand and ends with a line of asterisks Classification results table If table output is turned on Metaxa2 will save statistics of every BLAST match that the sequence in question produces against the database to in a file with the suffix blast table This file consists of tab separated columns containing information on the matches found one BLAST match per line The contents of the columns from left to right are explained in this table Column Description Query ID The identifier of the query sequence Subject ID The identifier of the matching database sequence Score The score this match has obtained in the classification system Species The species name of the database system Score The BLAST score of the match E value The E value of the match as reported by BLAST Each new query is indicated by a comment line e g Query AATT01000235 146421 147977 E List of non target sequences If not found output is turned on Metaxa2 will write a list of sequences for which no conserved SSU LSU regions could be found to a file with the suffix _not_found txt The file contains only the identifiers of the non rRNA sequences Sequence alignments By default Metaxa2 saves alignments of
26. ly origin that was found in the sequence The reliability score is calculated based on the per cent identity to the best BLAST hit and how divergent the rest of the BLAST hits are from the first one The maximal score is 100 and the minimum score is determined by the R cutoff used 80 by default Scores above 80 can generally be considered good Graphical representations Metaxa2 writes graphical ASCII representations of where in each sequence the various conserved regions were found to a text file with the suffix graph Separate graphs are written for each origin and strand which means that each sequence entry may be present more than once in this file if it have matches to HMM profiles from more than one origin This makes it possible to manually inspect how Metaxa2 has evaluated each sequence The graphical representations look like this B matches on main strand gt gt id 454_ 30 gi 50402825 gb AY687385 1 403 bp kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk The first row shows the type of the entries below as well as the strand they are found on Each entry begins with the characters gt gt followed by the sequence identifier and its length Below the identifier row the sequence graph is shown By default all sequences are scaled so that they are of equal length and the domains are placed according to their relatwe position in the sequence The characters that are used in the graphical repr
27. ly to version 1 Metaxa2 uses multiple HMM profiles representing conserved domains in the SSU sequence In addition separate sets of HMM profiles for SSU sequences of archaeal bacterial eukaryal mitochondrial and chloroplast origin are utilised We have also added HMM profiles representing the LSU gene in these lineages in very much the same way To avoid false positive matches Metaxa2 by default requires at least two such conserved domains to be found on a query sequence This criterion brings down the false positive rate to close to zero The subsequent BLAST based classification step now has an even more elaborately crafted and curated database allowing Metaxa2 to make accurate taxonomic predictions that often go down to genus or species levels The scoring system for assigning the sequences to archeal bacterial eukaryal mitochondrial or chloroplast origin still remains in Metaxa2 If the origin of the final classification does not agree with the predicted origin from the HMMER based step the sequence classification is marked as uncertain by applying a to the end of the definition line The sequence is also marked as uncertain if the difference between the scores of the two most likely origins is smaller than the number of sequences of analysed BLAST matches by default 5 Most changes that have been introduced in Metaxa2 are related to either the improved taxonomy engine or adaptions to modern metagenomics that is moving towa
28. mains for the most likely origin that was found in the sequence The average E value for these domains The average score for these domains The starting position of the first domain The ending position of the last domain The domain that is located first on the sequence The domain that is located last on the sequence The word Chimeric if the sequence was marked as a potential chimera Empty if not Sequences will only be marked as chimeric if the allow_reorder option is turned off Note that this is not a robust measure against chimeras of all kinds A collection of information of all possible origins for the given query Each entry is a space separated list containing the origin type the number of domains of that type the average E value and the average score e g N 4 8 2e 11 43 475 Extraction results table If table output is turned on Metaxa2 will save statistics of every profile set that the sequence in question matches to in a file with the suffix hmmer table This file consists of tab separated columns containing information on the rRNA sequence found The contents of the columns from left to right are explained in this table Column ID Length List of hits Description The identifier of the query sequence The length of the query sequence Each new column contains information of a profile match Each column is organised as follows lt starting position gt lt ending position gt
29. nd 25S or 28S rRNA in eukaryotes Toggling the switch from the SSU gene to the LSU gene is done using the g lsu or gene lsu options both set to ssu by default The operation of Metaxa 2 is then the same as when searching for and extracting SSU genes Metaxa s databases representing the SSU and LSU genes are kept separately within the metaxa2_db directory Although Metaxa2 s support for LSU extractions should be reliable and robust it has not been as extensively tested internally as the SSU extractions Therefore we encourage users to report suspicious or obviously misclassified entries so that Metaxa s LSU support can be improved even further in the future Completely redesigned system for taxonomic classifications One of the design goals of Metaxa2 was to be able to make more sensible predictions of the origin of each input sequence identified as an SSU or LSU gene even at very short read lengths To achieve this we have completely rewritten the classification engine which has enabled Metaxa2 to deliver reliable predictions of the taxonomic lineage of the extracted SSU LSU genes Metaxa reports those lineages to a file with the suffix taxonomy txt By using the taxonomic data of the best five BLAST matches by default Metaxa2 calculates the most accurate taxonomic placement possible for the input sequence given the lineages of the matching hits and their degree of identity to the input sequence Each
30. nput but separated for the different origins Off F by default unknown T or F Outputs a list of entries designated as unknowns with their statistics Off F by default Information options h Displays the help message help Displays the help message bugs Displays the bug fixes and known bugs in this version of Metaxa license Displays licensing information The Metaxa Taxonomic Traversal Tool is supposed to be run after Metaxa has completed and uses the taxonomy txt output file as its input 6 Internal changes in Metaxa2 The original algorithms and designs behind Metaxa are described in the manual for Metaxa version 1 1 2 see part 4 Algorithm and implementation in that manual This manual will mainly focus on the changes made from previous Metaxa versions instead of reiterating the information from the previous manual If the main design goal for the original version of Metaxa was to achieve fast and accurate extraction of SSU sequences in large data sets without introducing a large number of false positives the design goal of Metaxa2 is to do the same task faster better and for larger data sets In addition Metaxa2 adds the ability to extract and classify LSU sequences as well as more precise taxonomic classifications Metaxa still relies on the HMMER3 software which allows for extremely fast comparisons filtration of the input dataset and further analysis of just a subset of the input data Similar
31. ocated in the same directory as Metaxa itself The BLAST database used for classification By default Metaxa assumes to find the databases in the metaxa_db directory located in the same directory as Metaxa itself Adds a date and time stamp to the output file This can be useful e g if Metaxa is part of a pipeline where input files with the same name could cause overwriting of important data Off F by default If enabled the BLAST search will be performed using BLAST instead of the legacy blastall engine By default blastall is used F FASTQ and Paired end specific options Note that FASTQ format is only supported when Metaxaz is run in pipeline mode q value quality_percent value quality_filter T or F quality_trim T or F ignore_paired_read T or F distance value Minimum quality value for a nucleotide to be considered good Default 20 Percentage of low quality below q value accepted before filtering trimming is started Default 10 If enabled Metaxa2 will filter out low quality reads below specified q value removing them entirely from the input data before running the HMMER searches Off F by default If enabled Metaxa2 will trim bad bases below q value from the end of the read This is generally to prefer over filtering the read entirely Off F by default When enabled Metaxa2 will not discard the entire read pair if only one of the reads is of bad quality
32. rds even larger datasets but still with short read lengths In general the internal changes are invisible to the end user but some of them might have impact on particular usage scenarios First of all the main metaxa2 program now is able to perform quality filtering of sequences in FASTQ format Please note that the filtering is performed before metaxa2_x or metaxa2_c has been started Thus none of the individual programs can read FASTQ format Instead they still expect input in FASTA format Metaxa accommodate for paired end reads in a special fashion Similarly to FASTQ input paired end libraries must be pre processed through the first metaxa2 program and cannot be directly inputted into metaxa2_x or metaxa2_c This is because Metaxa2 re organizes the input sequences into a special concatenated format which enable HMMER to work on both ends at once This conversion however must happen before the extraction step can begin This design decision was taken to straight out the downstream processing of sequence entries Although it would be possible to re write the code so that the individual tools handle FASTQ and paired end input it would severely complicate the way Metaxa2 internally manages its files When Metaxa reads in a paired end library it first concatenates the two reads into one sequence retaining information on where the original reads begin and end The paired end information is saved to a file called pairinfo 1 txt which c
33. t of the respective program To get these file use the save_raw T see below Off F by default Table format output of probable taxonomic origins for the identified SSU LSU sequences File suffix is taxonomy txt On T by default Forces Metaxa to classify sequences at a certain taxonomy level regardless of their reliability score Off 0 by default If on Metaxa outputs a list of entries that do not seem to be SSU sequences File suffix is _not_found txt Off F by default Outputs alignments of BLAST matches to each query in all a uncertain u or no n cases Requires MAFFT to be installed Default is to output alignments in uncertain cases u Removes ends of SSU sequences if they are outside of the SSU region If off the whole input sequence is saved On T by default Depreciated option use the taxonomy option instead Kept for compatibility with previous Metaxa versions Off F by default silent T or F Suppresses printing of progress info to screen Off F by default graph_scale value Sets the scale of the graphical output If the provided value is zero a percentage view is shown Default is 0 save_raw T or F Saves all raw data for searches etc instead of removing it when finished Saves data to a directory with the suffix _metaxa_raw_output Off F by default Information options h Displays basic usage help help Displays the help message complete with a
34. tool outputs the number of identified SSU LSU sequences associated with each node in the taxonomic tree at different levels roughly corresponding to kingdoms phyla classes orders families genera species subspecies etc The guess_species option is now obsolete Since Metaxa now handles taxonomy in a much more elaborate way the guess_species option has been deemed obsolete It is however still useful but a warning message will appear The function may be depreciated in future versions of Metaxa and is not actively maintained anymore and has thus not been tested in the Metaxa testing Updated database for classification The improved classification engine has naturally required us to update and improve upon the underlying databases The basis of the Metaxa databases is the SILVA reference release 111 and Mitozoa release 10 From this data we have created a curated reference database in large part by automated means but also using extensive manual curation and cross checking in GreenGenes CRW and GenBank Suspicious or erroneous entries were removed as well as sequences from uncultured or unverified organisms Direct input of sequences in FASTQ format Metaxa2 now supports input of sequence data sets directly in the FASTQ format Although Metaxa will try to auto detect the format of the input file specifying f fastq will force Metaxa to read input in FASTQ format This option might be particularly usef
35. ul when piping input to Metaxa2 using standard input Note that FASTQ files are only supported when Metaxa2 is run in pipeline mode i e not when the two tools metaxa2_x and metaxa2_c are run separately These two tools still expect input in FASTA format Support for paired end libraries In addition to FASTQ support Metaxa2 also handles paried end libraries Paired end data needs to be supplied in two separate files with sequence in both files corresponding to the two ends of the same read and so on If your data is held in a single file a utility such as the pefcon and pesort tools of the PETKit http microbiology se software petkit can help you To enable Metaxa2 s paired end capabilities use the options 1 firstfile 2 secondfile instead of the i option for input Metaxa will automatically orient the reads in the same direction and utilize a combination of the two ends for the extraction and classification process Note that the two ends will be outputted together as one single read with a spacer Improved support for libraries with short read lengths 100 bp Since metagenomics is moving towards shorter read lengths and larger data sets e g Illumina and IonTorrent sequencing special care has been taken to try to address the short read issues in previous Metaxa versions We have tried to optimize the default options for reads down to 100 bp however this will increase the likelihood of false positive extractions
Download Pdf Manuals
Related Search