Skip to content

Latest commit

 

History

History
349 lines (269 loc) · 33.6 KB

README.md

File metadata and controls

349 lines (269 loc) · 33.6 KB
Raw

European Galaxy server GitHub release (latest SemVer) Conda PyPI version GitHub issues Docker Pulls GitHub commit activity

ECTyper (an easy typer)

ECTyper is a standalone versatile serotyping module for Escherichia coli. It supports both fasta (assembled) and fastq (raw reads) file formats. The tool provides convenient species identification coupled to quality control module giving a complete, transparent and reference laboratories suitable report on E.coli serotyping, Shiga toxin typing and pathotyping.

Introduction

Escherichia coli is a priority foodborne pathogen of public health concern and popular model organism. Phenotypic characterization such as serotyping, toxin typing and pathotyping provide critical information for surveillance and outbreak detection activities and research including source attribution, outbreak cluster assignment, pathogenicity potential, risk assessement and others.

ECTyper uses whole-genome sequencing (WGS) for E.coli characterization including species identification, in silico serotyping covering O and H antigens, Shiga toxin typing and DEC pathotyping. It is a versatile, scallable, easy to use tool allowing to obtain key information on E.coli accepting both raw and assembled inputs.

As WGS becomes standard within public health and research laboratories, it is important to harness the high throughput and resolution potential of this technology providing accurate and rapid at scale typing of E.coli both in public health, clinical and research contexts.

Citation

If you find ECTyper useful, please cite the following paper:

Bessonov, Kyrylo, Chad Laing, James Robertson, Irene Yong, Kim Ziebell, Victor PJ Gannon, Anil Nichani, Gitanjali Arya, John HE Nash, and Sara Christianson. "ECTyper: in silico Escherichia coli serotype and species prediction from raw and assembled whole-genome sequence data." Microbial genomics 7, no. 12 (2021): 000728. https://www.microbiologyresearch.org/content/journal/mgen/10.1099/mgen.0.000728

Contact

For any questions, issues or comments please make a Github issue or reach out to Kyrylo Bessonov.

Installation

Multiple installation options are available depending on the user context and needs. The most convinient installation is as a conda package as it will install all required dependencies.

Docker and Singularity images availability

Docker and Singularity images are also available from https://biocontainers.pro/tools/ectyper that could be useful for NextFlow or hassle-free deployment

Databases

ECTyper uses multiple databases

Option 1: As a conda package

Optionally if you do not have a conda environment, get and install miniconda or anaconda:

wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh
bash miniconda.sh -b -p $HOME/miniconda
echo ". $HOME/miniconda/etc/profile.d/conda.sh" >> ~/.bashrc
source ~/.bashrc

Install the latest ectyper conda package from a bioconda channel

conda install -c bioconda ectyper

Option 2: Install using pip

Install using pip3 utility including python but missing on non-python dependencies

pip3 install ectyper

Option 3: From source code

Second option is to install from the source allowing to excercise maximum control over installation process.

Install dependencies. On Ubuntu distro run

apt install samtools bowtie2 mash bcftools ncbi-blast+ seqtk

Install python dependencies via pip:

pip3 install pandas biopython

Clone the repository or checkout a particular release (e.g v1.0.0, v2.0.0 etc.):

git clone https://github.com/phac-nml/ecoli_serotyping.git
git checkout v2.0.0 #optionally checkout a specific release version

Finally, install ectyper from source

python3 setup.py install # option 1
pip3 install .   # option 2

Compatibility

Dependencies:

  • python >= 3.5
  • bcftools >= 1.8
  • blast == 2.7.1
  • seqtk >= 1.2
  • samtools >= 1.8
  • bowtie2 >= 2.3.4.1
  • mash >= 2.0

Python packages:

  • biopython >= 1.70
  • pandas >= 0.23.1
  • requests >= 2.0

Getting started

Basic Usage

  1. Put the fasta/fastq files for serotyping analyses in one folder (concatenate paired raw reads files if you would like them to be considered a single entity)
  2. ectyper -i [file path] -o [output_dir]
  3. View the results on the console or in cat [output folder]/output.csv

Example Input Scenarios

  • ectyper -i ecoliA.fasta for a single file (the output folder will be named using ectyper_<date>_<time> pattern as -o is not specified)
  • ectyper -i ecoliA.fasta -o output_dir for a single file with results stored in output_dir folder
  • ectyper -i ecoliA.fasta ecoliB.fastq ecoli_folder/ for multiple files and the directory ecoli_folder/ separated by space. Again the output folder will be named using ectyper_<date>_<time> pattern)
  • ectyper -i ecoliA.fasta ecoliB.fastq,ecoliC.fna for multiple files separated by comma , symbol
  • ectyper -i ecoli_folder scan for input files in a folder and subdirectories (all files in the folder will be checked by the tool)
  • ectyper -i ecoli_folder/*.fasta scan for FASTA input files in a folder and subdirectories with fasta extension

Advanced Usage

usage: ectyper [-h] [-V] -i INPUT [-c CORES] [-opid PERCENTIDENTITYOTYPE] [-hpid PERCENTIDENTITYHTYPE] [-opcov PERCENTCOVERAGEOTYPE] [-hpcov PERCENTCOVERAGEHTYPE] [--verify] [-o OUTPUT] [-r REFSEQ] [-s] [--debug] [--dbpath DBPATH]


ectyper v2.0.0 database v1.0 Prediction of Escherichia coli serotype from raw reads or assembled
genome sequences. The default settings are recommended.


options:
  -h, --help            show this help message and exit
  -V, --version         show program's version number and exit
  -i INPUT, --input INPUT
                        Location of E. coli genome file(s). Can be a single file, a comma-
                        separated list of files, or a directory
  --longreads           Enable for raw long reads FASTQ inputs (ONT, PacBio, other platforms) [default False]                      

  -c CORES, --cores CORES
                        The number of cores to run ectyper with
  -opid PERCENTIDENTITYOTYPE, --percentIdentityOtype PERCENTIDENTITYOTYPE
                        Percent identity required for an O antigen allele match [default 95]
  -hpid PERCENTIDENTITYHTYPE, --percentIdentityHtype PERCENTIDENTITYHTYPE
                        Percent identity required for an H antigen allele match [default 95]
  -opcov PERCENTCOVERAGEOTYPE, --percentCoverageOtype PERCENTCOVERAGEOTYPE
                        Minumum percent coverage required for an O antigen allele match [default 90]
  -hpcov PERCENTCOVERAGEHTYPE, --percentCoverageHtype PERCENTCOVERAGEHTYPE
                        Minumum percent coverage required for an H antigen allele match [default 50]
  --verify              Enable E. coli species verification
  -o OUTPUT, --output OUTPUT
                        Directory location of output files
  -r REFSEQ, --refseq REFSEQ
                        Location of pre-computed MASH RefSeq sketch. If provided, genomes identified as non-E. coli will have their species identified using MASH. For best results the pre-sketched RefSeq archive https://gembox.cbcb.umd.edu/mash/refseq.genomes.k21s1000.msh is recommended
  -s, --sequence        Prints the allele sequences if enabled as the final columns of the output
  --debug               Print more detailed log including debug messages
  --dbpath DBPATH       Path to a custom database of O and H antigen alleles in JSON format. Check Data/ectyper_database.json for more information

Configuration and fine-tunning parameters

ECTyper requires minimum options to run (-i input directory or file) but allows for extensive configuration to accomodate wide variaty of typing scenarios

Parameter Explanation Usage scenario
--maxdirdepth Maximum number of directory levels to use in a directory for input file search Given a directory path use this number of levels (default 0 levels) to go down the specified directory path in search for files. Useful is a directory contains multiple sub-directories
-c The number of cores to run on Use multiple cores to run multiple samples and subtools on. Especially useful if a directory contains multiple files or working with raw reads in FASTQ format
-opid Minimum %identity threshold just for O antigen match Poor coverage of O antigen genes or for exploratory work (recommended value is 95)
-opcov Minimum %covereage threshold for a valid match against reference O antigen alleles Poor coverage of O antigen genes and a user wants to get O antigen call regardless (recommend value is 90)
-hpid Minimum %identity threshold just for H antigen match Poor coverage of O antigen genes or for exploratory work (recommend value is 95)
-hpcov Minimum %covereage threshold for a valid match against reference H antigen alleles Poor coverage of O antigen genes and a user wants to get O antigen call regardless (recommend value is 50)
--verify Verify species of the input and run QC module providing information on the reliability of the result and any typing issues User not sure if sample is E.coli and wants to obtain if serotype prediction is of sufficient quality for reporting purposes and predicted species is E.coli. Note that pathotyping and Shiga toxin module will not run if --verify is specified and species is other than E.coli
-r Specify custom MASH sketch of reference genomes that will be used for species inference User has a new assembled genome that is not available in NCBI RefSeq database. Make sure to add metadata to assembly_summary_refseq.txt and provide custom accession number that start with GCF_ prefix
--dbpath Provide custom appended database of O and H antigen reference alleles in JSON format following structure and field names as default database ectyper_alleles_db.json User wants to add new alleles to the alleles database to improve typing performance
--pathotype Perform DEC pathotype and Shiga toxin subtyping prediction on a sample Predict DEC pathotype using the key diagnostic pathotype markers. Results are only valid for E.coli samples. Also performs Shiga toxin subtyping if stx1 or stx2 genes are present
-pathpid Minimum %identity threshold for pathotype and Shiga toxin subtyping results filtering All gene hits from pathotype database also containing stx genes will use this minimum cutoff to filter results. The default value is minimum 90% identity
-pathcov Minimum %coverage threshold for pathotype and Shiga toxin subtyping results filtering All gene hits from pathotype database also containing stx genes will use this minimum cutoff to filter results. The default value is minimum 50% coverage

Data Input

Both raw and assembled reads are accepted in FASTA and FASTQ formats from any sequencing platform. The tool was designed for single sample inputs, but was shown to work on multi-taxa metagenomic raw reads FASTQ inputs.

Data Output

The output of the tool is stored in text files with the main report stored in output.tsv tab-delimited text file.

The BLASTN hits of the O and H antigen database are stored in blastn_output_alleles.txt tab-delimited file.

The log messages are stored in ectyper.log text file

{out folder name}
├── blastn_output_alleles.txt
├── ectyper.log
└── output.tsv

Main submodules descriptions

Species identification module

ECTyper performs species identification by selecting the closest reference genome to the query input (i.e.having the smallest MASH distance) from the custom made MASH sketch currently represented by the 119,980 reference genomes with assigned taxonomy information. The enhanced ECTyper species ID sketch is both based on The Genome Taxonomy Database (GTDB) release 214 (covering all known bacteria and archaea domains) and manually curated and selected Escherichia and Shigella genera reference genomes from the EnteroBase and GenBank/RefSeq NCBI public databases.

The comment field of the MASH sketch uses taxonomic path from domain to species level to meet the MikroKondo MASH sketch guidelines. For example, taxonomic path for E.coli is formatted as such "d__Bacteria;p__Pseudomonadota;c__Gammaproteobacteria;o__Enterobacterales;f__Enterobacteriaceae;g__Escherichia;s__Escherichia coli". The sketch was generated using 1000 hashes per sketch (-s 1000) and k-mer size of 21 (-k 21) that are the default mash sketch parameters.

The Escherichia and Shigella genera genomes sourced from EnteroBase accessed on 2023-09-07 were assembled using the shovill v1.1.0 assembler run on raw reads downloaded from the NCBI SRA via the corresponding SRA accession number. The Escherichia and Shigella RefSeq NCBI genomes were downloaded from NCBI FTP by accessing metadata from the assembly_summary_refseq.txt file, filtering the organism_name field using all known Escherichia and Shigella genera species. The reported species for the EnteroBase and RefSeq sourced genomes were verified by running the ShigEiFinder v1.3.5 and checking the SEROTYPE field. Samples with values SB13 (Shigella boydii type 13) and Unknown were rejected.

For each Escherichia and Shigella species selected genomes a distance matrix was calculated by running mash triangle followed by the agglomerative hierarchical clustering using average linkage method via the linkage() function form SciPy package. Next flat clusters were formed on the previously calculated hierarchical clustering object using the SciPy fcluster() function with distance as the clusters forming criterion. The flat clusters with 2 or more members (i.e. genomes) were selected and clusters with a single member (i.e. singletons) were discarded as potential outliers or noise. Finally the cluster centroids (i.e. representative cluster genomes) were defined by the partition around medoids algorithm (PAM) via the fasterpam() function from the kmedoids package. These cluster centroids formed the final list of Escherichia and Shigella genomes selected for the original GTDB sketch enhancement. Currently the MASH ECTyper species sketch contains 34,736 Escherichia and 4,627 Shigella genomes.

The ECTyper species identification module performance was tested and validated against the highly curated 493 genomes with species and cgMLST clustering data covering Shigella flexneri, Shigella dysenteriae, Shigella boydii, Shigella sonnei and E.coli species described by the Iman Yassine et al. 2022

Serotyping module

Independent of the input type (FASTQ or FASTA) the reference database of O and H antigen alleles is being scanned. The best matching alleles hits for O antigen (wzx, wzy, wzm, wzt) and H antigen (mostly represented by fliC in addition to flkA, fllA and flmA) genes are ranked by maximizing BOTH %identity and %coverage values via the allele gene score=(%identity*%coverage)/10000 using the BLASTN default run parameters. For O and H antigen the highest scoring antigen is reported. Since some O antigens display very high level of sequence similarity represented by the 16 high similarity O groups delineated by the Atsushi Iguchi et al. 2015, the mixed O antigen calls are possible separated by the \ symbol such as the group 9 that will be reported as O17/O44/O73/O77/O106 mixed O serotype call. To account for any sequencing and other errors, if predicted O antigen is a member of the high similarity group, all O antigens in the group will be reported resulting in a mixed O antigen final call. The H antigens are better separated and are always reported as a single antigen call. For more serotyping details and benchmarking results please see the ECTyper v1.0.0 publication in the Citation section.

Pathotyping module

Most of the E.coli samples are non-pathogenic, but some present public health risk and could lead from mild to sever health conditions. ECTyper currently supports typing of the 7 diarrheagenic Escherichia coli (DEC) pathotypes: DAEC, EAEC, EHEC, EIEC, EPEC, ETEC and STEC. Please note that the EHEC pathotype is being reported as EHEC-STEC as it is a STEC subtype that could cause bloody dysentery and hemolytic uremic syndrome (HUS).

This module uses highly curated database of the diagnostic pathotype markers listed in the pathotype and toxin typing database in JSON format (see Databases section). The database was assembled by curating the existing literature and tools. Each database entry contains diagnostic marker accession number, nucleotide sequence and its length, gene symbol, source and other information in the following format:

{
  "accession": "QGG66503.1",
  "description": "intimin",
  "dnasequence": "ATGATTACTCATGGTTTTT...",
  "gene": "eae",
  "protsequence": "MITHGFYARTRHKHKLK...",
  "source": "https://github.com/B-UMMI/patho_typing/",
  "subtype": "-",
  "id": "28",
  "length_dnasequence": "2820"
}

Importantly the pathotype database under the pathotypes key contains the set of pathotyping classification gene-based rules that are used to predict a given E.coli pathotype from BLASTN output using the following format below.

"EPEC": {
          "name": "Enteropathogenic Escherichia coli",
          "rules": {
          "16": {
                  "genes": [
                        "eae",
                        "!stx1",
                        "!stx2"
                    ],
                    "publication": "PMID:11403406"
                }
          }
}

The ! in front of the gene symbol (e.g., !stx2 ) indicates the negation that is a given gene must NOT BE PRESENT in order to satisfy a given rule.

Each pathotype classification rule might have one or more genes listed under the genes key. Each rule also has a corresponding publication reference under the publication key as an additional evidence and rule validity support.

Each rule is tested for presence or absence of genes listed under the genes rule key. If all presence or absence conditions (i.e. the genes marked by the !) are met a given rule is considered to be valid and pathotype assigned. Thus a given sample might have several rules that would concurrently apply that could lead to mixed pathotype prediction separated via the / symbol such as ETEC/STEC.

Shiga toxin typing module

The Shiga toxin subtyping module supports typing of the stx1 and stx2 gene subtypes that is relevant both for epidemiological and risk assessment purposes (e.g., disease severity). This module also heavily relies on the pathotype and toxin typing database (see Databases section).

Currently the database supports

  • 4 stx1 subtypes: stx1a, stx1c, stx1d and stx1e
  • 15 stx2 subtypes: stx2a, stx2b, stx2c, stx2d, stx2e, stx2f, stx2g ,stx2h, stx2i, stx2j, stx2k,stx2l, stx2m, stx2n, stx2o.

The input sequences are queried against the stx1 and stx2 markers via BLASTN and top hits are being reported separated by the ; symbol. The module supports the multi-copy stx gene presence by taking into account the genomic stx location attributes for each stx subtype (i.e. gene coordinates, contig location, overlap with other stx hits). The multi-copy stx gene reporting is not exhaustive (not all hits are being reported). That is if multiple stx hits are found in the input, the highest quality hit(s) per each non-overlapping stx gene range is being reported (i.e. single or multiple top hits are possible with the highest identical bitscore value as some hits could not be resolved due to sequence truncation). For example, if several stx allele hits have identical bitscore in a given stx gene range, all such hits are being reported. Note that the StxSubtypes field lists only UNIQUE stx subtypes for the entire input sample such as stx2e;stx2k even if their genomic locations overlap or are identical due to truncated incomplete stx allele signatures. The StxContigNames and StxCoordinates lists all contig names and corresponding genomic coordinates for each listed stx type in the StxSubtypes field according to the alphabetical order. This allows to easily spot stx subtypes with the same genomic coordinates. Finally, these fields allow to better understand stx alleles context/function and spot truncated alleles while providing genomic location context.

Quality Control (QC) module

To provide an easier interpretation of the results and typing metrics, following QC codes were developed. These codes allow to quickly filter "reportable" and "non-reportable" samples. The QC module is tightly linked to ECTyper allele database, specifically, MinPident and MinPcov fields. For each reference allele minimum %identity and %coverage values were determined as a function of potential "cross-talk" between antigens (i.e. multiple potential antigen calls at a given setting). The QC module covers the following serotyping scenarios. More scenarios might be added in future versions depending on user needs.

QC flag Explanation
PASS (REPORTABLE) Both O and H antigen alleles meet min %identity or %coverage thresholds (ensuring no antigen cross-talk) and single antigen predicted for O and H
FAIL (-:- TYPING) Sample is E.coli and O and H antigens are not typed. Serotype: -:-
WARNING MIXED O-TYPE A mixed O antigen call is predicted requiring wet-lab confirmation
WARNING (WRONG SPECIES) A sample is non-E.coli (e.g. E.albertii, Shigella, etc.) based on RefSeq assemblies
WARNING (-:H TYPING) A sample is E.coli and O antigen is not predicted (e.g. -:H18)
WARNING (O:- TYPING) A sample is E.coli and O antigen is not predicted (e.g. O17:-)
WARNING (O NON-REPORT) O antigen alleles do not meet min %identity or %coverage thresholds
WARNING (H NON-REPORT) H antigen alleles do not meet min %id or %cov thresholds
WARNING (O and H NON-REPORT) Both O and H antigen alleles do not meet min %identity or %coverage thresholds

Report format (output.tsv)

ECTyper capitalizes on a concise minimum output coupled to easy results interpretation and reporting. ECTyper v1.0 serotyping results are available in a tab-delimited output.tsv file consisting of the 16 columns listed below:

  1. Name: Sample name (usually a unique identifier)
  2. Species: the species column provides valuable species identification information in case of inadvertent sample contamination or mislabelling events
  3. O-type: O antigen
  4. H-type: H antigen
  5. Serotype: Predicted O and H antigen(s)
  6. QC: The Quality Control value summarizing the overall quality of prediction
  7. Evidence: How many alleles in total used to both call O and H antigens
  8. GeneScores: ECTyper O and H antigen gene scores in 0 to 1 range
  9. AllelesKeys: Best matching ECTyper database allele keys used to call the serotype
  10. GeneIdentities(%): %identity values of the query alleles
  11. GeneCoverages(%): %coverage values of the query alleles
  12. GeneContigNames: the contig names where the query alleles were found
  13. GeneRanges: genomic coordinates of the query alleles
  14. GeneLengths: allele lengths of the query alleles
  15. Database: database release version and date
  16. Warnings: any additional warnings linked to the quality control status or any other error message(s).

Selected columns from the ECTyper typical serotyping report

Name Species Serotype Evidence QC GeneScores AlleleKeys GeneIdentities(%) GeneCoverages(%) GeneContigNames GeneRanges GeneLengths Database Warnings
15-520 Escherichia coli O174:H21 Based on 3 allele(s) PASS (REPORTABLE) wzx:1; wzy:1; fliC:1; O104-5-wzx-origin;O104-13-wzy;H7-6-fliC-origin; 100;100;100; 100;100;100; contig00049;contig00001;contig00019; 22302-23492;178-1290;6507-8264; 1191;1113;1758; v1.0 (2020-05-07) -
EC20151709 Escherichia coli O157:H43 Based on 3 allele(s) PASS (REPORTABLE) wzx:1;wzy:0.999;fliC:1 O157-5-wzx-origin;O157-9-wzy-origin;H43-1-fliC-origin; 100;99.916;99.934; 100;100;100; contig00002;contig00002;contig00003; 62558-63949;64651-65835;59962-61467; 1392;1185;1506; v1.0 (2020-05-07) -

Selected columns from the ECTyper typical pathotyping and shiga toxin subtyping report

Name Species SpeciesMashRatio SpeciesMashTopID Serotype Pathotype PathotypeGenes PathotypeGeneNames PathotypeAccessions PathotypeIdentities(%) PathotypeCoverages(%) PathotypeRuleIDs PathotypeGeneCounts StxSubtypes StxAccessions StxIdentities(%) StxCoverages(%) StxLengths StxContigNames StxCoordinates
SRR7947260 Escherichia coli 992/1000 GCF_005046225.1 O109:H16 ETEC/STEC ehxA,hlyE,sta1,stx2 ehxA: enterohemolysin A,hlyE: hemolysin E chromosomal,sta1: heat-stable enterotoxin ST,stx2: Shiga toxin 2 subtype a complete sequence (subunits A and B) EF204927.1;AVXX01000010.1;AJ555214;AB030484 99.619;97.807;95.89;100.0 70;100;100;100 20:ETEC:sta1;27:STEC:stx2 ETEC:1 (sta1);STEC:1 (stx2) stx2a;stx2g AB030484;AY286000 100.0;99.758 100;100 1241;1242 contig00064;contig00074 3723-4963;14725-15966

FAQ

Does ECTyper can be run on multiple samples in a directory?

ECTyper proves flexible ways to specify inputs located in different locations. One can provide multiple paths to several directories separated by space. In addition, one can specify file type to look for in a given directory(ies). Note that paths that contain a star * symbol would only look for files in specified directory and would not look in subdirectories. For example,

  • Process all files in folder1 and folder2 directories and file sample.fasta located in folder3

    ectyper -i folder1/ folder2/ folder3/sample.fasta -o ectyper_results

  • Process all fasta files in folder1 and all fastq files in folder2. All sub-directories in those 2 folders will be ignored. To process those sub-folders either specify path to them or provide paths to directories without the * wildcard symbol.

    ectyper -i folder1/*.fasta folder2/*.fastq

Why ECTyper sometimes provides serotype results separated by forward slash / for O-antigen

Some O-antigens display very high degree of homology and are very hard to discern even using wet-lab agglutination assays. Even using both wzx and wzy genes it is not possible to reliably resolve those O-antigens. The 16 high similarity groups were identified by Joensen, Katrine G., et al.. Thus, if a given O-antigen is a member of any of those high similarity groups, all potential O-antigens are reported separated by / such as group 9 reported as O17/O44/O73/O77/O106.

Availability

Resource Description Type
PyPI PyPI package that could be installed via pip utility Terminal
Conda Conda package available from BioConda channel Terminal
Docker Images containing completely initialized ECTyper with all dependencies Terminal
Singluarity Images containing completely initialized ECTyper with all dependencies Terminal
GitHub Install source code as any Python package Terminal
Galaxy ToolShed Galaxy wrapper available for installation on a private/public instance Web-based
Galaxy Europe Galaxy public server to execute your analysis from anywhere Web-based
IRIDA plugin IRIDA instances could easily install additional pipeline Web-based

Legal and Compliance Information

Copyright Government of Canada 2024

Written by: National Microbiology Laboratory, Public Health Agency of Canada

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this work except in compliance with the License. You may obtain a copy of the License at:

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.