EsMeCaTa: Estimating Metabolic Capabilties from Taxonomic annotations

EsMeCaTa is a tool to estimate metabolic capabilities from a taxonomic annotations (for example after analysis on 16S RNA sequencing). This is useful if no sequenced genomes or proteomes are available.

Warning: EsMeCaTa is in development so it can be unstable.

Table of contents

• EsMeCaTa: Estimating Metabolic Capabilties from Taxonomic annotations
  • Table of contents
  • Requirements
  • Installation
  • Input
  • EsMeCaTa commands
  • EsMeCaTa functions
    • esmecata proteomes: Retrieve proteomes associated to taxonomic annotation
    • esmecata clustering: Proteins clustering
    • esmecata annotation: Retrieve protein annotations
  • EsMeCaTa outputs
    • EsMeCaTa proteomes
    • EsMeCaTa clustering
    • EsMeCaTa annotation

Requirements

EsMeCaTa needs the following python packages:

• biopython: To create fasta files.
• pandas: To read the input files.
• requests: For the REST queries on Uniprot.
• ete3: To analyse the taxonomic annotations and extract taxon_id, also used to deal with taxon associated with more than 100 proteomes.
• SPARQLwrapper: Optionnaly, you can use SPARQL queries instead of REST queries. This can be done either with the Uniprot SPARQL Endpoint (with the option --sparql uniprot) or with a Uniprot SPARQL Endpoint that you created locally (it is supposed to work but not tested, only SPARQL queries on the Uniprot SPARQL endpoint have been tested). Warning: using SPARQL queries will lead to minor differences in functional annotations and metabolic reactions due to how the results are retrieved with REST query or SPARQL query.

Also esmecata requires mmseqs2 for protein clustering:

• mmseqs2: To cluster proteins.

Installation

The easiest way to install the dependencies of EsMeCaTa is by using conda:

conda install mmseqs2 pandas sparqlwrapper requests biopython ete3

A conda package for esmecata will be created in the future.

EsMeCata can be installed with pip command (in esmecata directory):

git clone https://github.com/ArnaudBelcour/esmecata.git

cd esmecata

pip install -e .

Input

EsMeCaTa takes as input a tabulated or an excel file with two columns one with the ID corresponding to the taxonomic annotation (for example the OTU ID for 16S RNA sequencing) and a second column with the taxonomic classification separated by ';'. In the following documentation, the first column (named observation_name) will be used to identify the label associated to each taxonomic annotation. An example is located in the test folder (Example.tsv).

For example:

It is possible to use EsMeCaTa with a taxonomic annotation containing only one taxon:

But this can cause issue. For example, "Cluster_2" is associated to Yersinia but two genus are associated to this name (one mantid (taxId: 444888) and one bacteria (taxId: 629)). EsMeCaTa will not able to differentiate them. But if you give more informations by adding more taxons (for example: 'Bacteria;Gammaproteobacteria;Yersinia'), EsMeCaTa will compare all the taxons of the taxonomic annotation (here: 2 (Bacteria) and 1236 (Gammaproteobacteria)) to the lineage associated to the two taxIDs (for bacteria Yersinia: [1, 131567, 2, 1224, 1236, 91347, 1903411, 629] and for the mantid one: [1, 131567, 2759, 33154, 33208, 6072, 33213, 33317, 1206794, 88770, 6656, 197563, 197562, 6960, 50557, 85512, 7496, 33340, 33341, 6970, 7504, 7505, 267071, 444888]). In this exmaple, there is 2 matches for the bacteria one (2 and 1236) and 0 for the mantid one. So EsMeCaTa will select the taxId associated to the bacteria (629).

EsMeCaTa commands

usage: esmecata [-h] [--version] {proteomes,clustering,annotation} ...

From taxonomic annotation to metabolism using Uniprot. For specific help on each subcommand use: esmecata {cmd} --help

optional arguments:
  -h, --help            show this help message and exit
  --version             show program's version number and exit

subcommands:
  valid subcommands:

  {proteomes,clustering,annotation}
    proteomes           Download proteomes associated to taxon from Uniprot Proteomes.
    clustering          Cluster the proteins of the different proteomes of a taxon into a single set of representative
                        shared proteins.
    annotation          Retrieve protein annotations from Uniprot.

Requires: mmseqs2 and an internet connection (for REST and SPARQL queries, except if you have a local Uniprot SPARQL endpoint).

EsMeCaTa functions

esmecata proteomes: Retrieve proteomes associated to taxonomic annotation

usage: esmecata proteomes [-h] -i INPUT_FILE -o OUPUT_DIR [-b BUSCO] [--ignore-taxadb-update] [--all-proteomes] [-s SPARQL] [--remove-tmp] [-l LIMIT_MAXIMAL_NUMBER_PROTEOMES] [--beta] [-r RANK_LIMIT]

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT_FILE, --input INPUT_FILE
                        Input taxon file (excel, tsv or csv) containing a column associating ID to a taxonomic annotation (separated by ;).
  -o OUPUT_DIR, --output OUPUT_DIR
                        Output directory path.
  -b BUSCO, --busco BUSCO
                        BUSCO percentage between 0 and 1. This will remove all the proteomes without BUSCO score and the score before the selected ratio of completion.
  --ignore-taxadb-update
                        If you have a not up-to-date version of the NCBI taxonomy database with ete3, use this option to bypass the warning message and use the old version.
  --all-proteomes       Download all proteomes associated to a taxon even if they are no reference proteomes.
  -s SPARQL, --sparql SPARQL
                        Use sparql endpoint instead of REST queries on Uniprot.
  --remove-tmp          Delete tmp files to limit the disk space used: files in tmp_proteome for esmecata proteomes and files created by mmseqs (in mmseqs_tmp).
  -l LIMIT_MAXIMAL_NUMBER_PROTEOMES, --limit-proteomes LIMIT_MAXIMAL_NUMBER_PROTEOMES
                        Choose themaximal number of proteomes after which the tool will select a subset of proteomes instead of using all the available proteomes (default is 99).
  --beta                Use uniprot beta REST query.
  -r RANK_LIMIT, --rank-limit RANK_LIMIT
                        This option limit the rank used by the tool for searching for proteomes. The given rank and all the superior ranks will be ignored. Look at the readme for more information (and a list of possible rank).

For each taxon in each taxonomic annotations EsMeCaTa will use ete3 to find the corresponding taxon ID. Then it will search for proteomes associated to these taxon ID in the Uniprot Proteomes database.

If there is more than 100 proteomes, esmecata will apply a specific method:

• (1) use the taxon ID associated to each proteomes to create a taxonomicc tree with ete3.

• (2) from the root of the tree (the input taxon), esmecata will find the direct deescendant (sub-taxons).

• (3) then esmecata will compute the number of proteomes associated to each sub-taxon.

• (4) the corresponding proportions will be used to select randomly a number of proteomes corresponding to the proportion.

For example: for the taxon Clostridiales, 645 proteomes are found. Using the organism taxon ID associated to the 645 proteomes we found that there is 17 direct sub-taxons. Then for each sub-taxon we compute the percentage of proportion of proteomes given by the sub-taxon to the taxon Clostridiales. There is 198 proteomes associated to the sub-taxon Clostridiaceae, the percentage will be computed as follow: 198 / 645 = 30% (if a percentage is superior to 1 it will be round down and if the percentage is lower than 1 it will be round up to keep all the low proportion sub-taxons). We will use this 30% to select randomly 30 proteomes amongst the 198 proteomes of Clostridiaceae. This is done for all the other sub-taxons, so we get a number of proteomes around 100 (here it will be 102). Due to the different rounds (up or down) the total number of proteomes will not be equal to exactly 100 but it will be around it. The number of proteomes leading to this behavior is set to 99 by default but the user can modify it with the -l/--limit-proteomes option.

Then the proteomes found will be downloaded. For protein with isoforms, the canonical sequence is retrieved except when the isoforms are separated in different Uniprot entries.

esmecata proteomes options:

• -s/--sparql: use SPARQL instead of REST requests

It is possible to avoid using REST queries for esmecata and instead use SPARQL queries. This option need a link to a sparql endpoint containing UniProt. If you want to use the SPARQL endpoint of UniProt, you can use the argument: -s uniprot.

• -b/--busco: filter proteomes using BUSCO score (default is 0.9)

It is possible to filter proteomes according to to their BUSCO score (from Uniprot documentation: The Benchmarking Universal Single-Copy Ortholog (BUSCO) assessment tool is used, for eukaryotic and bacterial proteomes, to provide quantitative measures of UniProt proteome data completeness in terms of expected gene content.). It is a percentage between 0 and 1 showing the quality of the proteomes that esmecata will download. By default esmecata uses a BUSCO score of 0.80, it will only download proteomes with a BUSCO score of at least 80%.

• --ignore-taxadb-update: ignore need to udpate ete3 taxaDB

If you have an old version of the ete3 NCBI taxonomy database, you can use this option to use esmecata with it.

• --all-proteomes: download all proteomes (reference and non-reference)

By default, esmecata will try to downlaod the reference proteomes associated to a taxon. But if you want to download all the proteomes associated to a taxon (either if they are non reference proteome) you can use this option. Without this option non-reference proteoems can also be used if no reference proteomes are found.

• --remove-tmp: remove proteomes stored in tmp_proteomes folder

• -l/--limit-proteomes: choose the number of proteomes that will lead to the used of the selection of a subset of proteomes

To avoid working on too many proteomes, esmecata works on subset of proteomes when there is too many proteomes (by default this limit is set on 99 proteomes). Using this option the user can modify the limit.

• -r/--rank-limit: This option limit the rank used by the tool for searching for proteomes. The given rank and all the superior ranks will be ignored.

To avoid working on rank with too much proteomes (which can have an heavy impact on the number of proteomes downloaded and then on the clustering) it is possible to select a limit on the taxonomic rank used by the tool.

The selected rank will be used to find the ranks to keep. For example, if the rank 'phylum' is given, all the rank below (from subphylum to isolate) will be kept. And the ranks from phylum to superkingdom will be ignored when searching for proteomes. The following ranks can be given to this option (from Supplementary Table S3 of PMC7408187):

Some ranks (which are not non-hierarchical) are not used for the moment when using this method (so some taxons can be removed whereas they are below a kept rank):

esmecata clustering: Proteins clustering

usage: esmecata clustering [-h] -i INPUT_DIR -o OUPUT_DIR [-c CPU] [-t THRESHOLD_CLUSTERING] [-m MMSEQS_OPTIONS] [--linclust] [--remove-tmp]

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT_DIR, --input INPUT_DIR
                        This input folder of clustering is the output folder of proteomes command.
  -o OUPUT_DIR, --output OUPUT_DIR
                        Output directory path.
  -c CPU, --cpu CPU     CPU number for multiprocessing.
  -t THRESHOLD_CLUSTERING, --threshold THRESHOLD_CLUSTERING
                        Proportion [0 to 1] of proteomes required to occur in a proteins cluster for that cluster to be kept in core proteome assembly.
  -m MMSEQS_OPTIONS, --mmseqs MMSEQS_OPTIONS
                        String containing mmseqs options for cluster command (except --threads which is already set by --cpu command and -v). If nothing is given, esmecata will used the option "--min-seq-id 0.3 -c 0.8"
  --linclust            Use mmseqs linclust (clustering in lienar time) to cluster proteins sequences. It is faster than mmseqs cluster (default behaviour) but less senstitive.
  --remove-tmp          Delete tmp files to limit the disk space used: files in tmp_proteome for esmecata proteomes and files created by mmseqs (in mmseqs_tmp).

For each taxon (a row in the table) EsMeCaTa will use mmseqs2 to cluster the proteins (using an identity of 30% and a coverage of 80%, these values can be changed with the --mmseqsoption). Then if a cluster contains at least one protein from each proteomes, it will be kept (this threshold can be changed using the --threshold option). The representative proteins from the cluster will be used. A fasta file of all the representative proteins will be created for each taxon.

esmecata clustering options:

• -t/--threshold: threshold clustering

It is possible to modify the requirements of the presence of at least one protein from each proteomes in a cluster to keep it. Using the threshold option one can give a float between 0 and 1 to select the ratio of representation of proteomes in a cluster to keep.

For example a threshold of 0.8 means that all the cluster with at least 80% representations of proteomes will be kept (with a taxon, associated with 10 proteomes, it means that at least 8 proteomes must have a protein in the cluster so the cluster must be kept).

• -c/--cpu: number of CPU for mmseqs2

You can give a numbe of CPUs to parallelise mmseqs2.

• -m/--mmseqs: mmseqs option to be used for the clustering.

String containing mmseqs options for cluster command (except --threads which is already set by --cpu command and -v). If nothing is given, esmecata will used the option "--min-seq-id 0.3 -c 0.8". For example you can give --mmseqs "--min-seq-id 0.8 --kmer-per-seq 80" to ask for a minimal identity between sequence of 80% and having 80 kmers per sequence.

• --linclust: replace mmseqs cluster by mmseqs linclust (faster but less sensitive)

Use mmseqs linclust (clustering in lienar time) to cluster proteins sequences. It is faster than mmseqs cluster (default behaviour) but less senstitive.

• --remove-tmp: remove mmseqs files stored in mmseqs_tmp folder

esmecata annotation: Retrieve protein annotations

usage: esmecata annotation [-h] -i INPUT_DIR -o OUPUT_DIR [-s SPARQL] [-p PROPAGATE_ANNOTATION] [--uniref] [--expression]

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT_DIR, --input INPUT_DIR
                        This input folder of annotation is the output folder of clustering command.
  -o OUPUT_DIR, --output OUPUT_DIR
                        Output directory path.
  -s SPARQL, --sparql SPARQL
                        Use sparql endpoint instead of REST queries on Uniprot.
  -p PROPAGATE_ANNOTATION, --propagate PROPAGATE_ANNOTATION
                        Proportion [0 to 1] of the reccurence of an annotation to be propagated from the protein of a cluster to the reference protein of the cluster. 0 mean the annotaitons from all proteins are propagated to the
                        reference and 1 only the annotation occurring in all the proteins of the cluster.
  --uniref              Use uniref cluster to extract more annotations from the representative member of the cluster associated to the proteins. Needs the --sparql option.
  --expression          Extract expresion information associated to the proteins. Needs the --sparql option.

For each of the representative proteins conserved, esmecata will look for the annotation (GO terms, EC number, function, gene name, Interpro) in Uniprot.

Then esmecata will create a tabulated file for each row of the input file and also a folder containg PathoLogic file that can be used as input for Pathway Tools.

esmecata annotation options:

• -s/--sparql: use SPARQL instead of REST requests

It is possible to avoid using REST queries for esmecata and instead use SPARQL queries. This option need a link to a sparql endpoint containing UniProt. If you want to use the SPARQL endpoint, you can just use: -s uniprot.

• -p/--propagate: propagation of annotation

It is possible to modify how the annotations are retrieved. By default, esmecata will take the annotations from the representative proteins. But with the -p option it is possible to propagate annotation form the proteins of the cluster to the reference proteins.

This option takes a float as input between 0 and 1, that will be used to filter the annotations retrieved. This number is multiplicated with the number of protein in the cluster to estimate a threshold. To keep an annotation the number of the protein having this annotaiton in the cluster must be higher than the threshold. For example with a threshold of 0.5, for a cluster of 10 proteins an annotation will be kept if 5 or more proteins of the cluster have this annotation.

If the option is set to 0, there will be no filter all the annotation of the proteins of the cluster will be propagated to the reference protein (it corresponds to the union of the cluster annotations). This parameter gives the higher number of annotation for proteins. If the option is set to 1, only annotations that are present in all the proteins of a cluster will be kept (it corresponds to the intersection of the cluster annotations). This parameter is the most stringent and will limit the number of annotations associated to a protein.

For example, for the same taxon the annotaiton with the parameter -p 0 leads to the reconstruction of a metabolic networks of 1006 reactions whereas the parameter -p 1 creates a metabolic network with 940 reactions (in this example with no use of the -p option, so without annotaiton propagation, there was also 940 reacitons inferred).

• --uniref: use annotation from uniref

To add more annotations, esmecata can search the UniRef cluster associated to the protein associated to a taxon. Then the representative protein of the cluster will be extracted and if its identity with the protein of interest is superior to 90% esmecata will find its annotaiton (GO Terms and EC numbers) and will propagate these annotations to the protein. At this moment, this option is only usable when using the --sparql option.

• --expression: extract expression information

With this option, esmecata will extract the expression information associated to a protein. This contains 3 elements: Induction, Tissue specificity and Disruption Phenotype. At this moment, this option is only usable when using the --sparql option.

EsMeCaTa outputs

EsMeCaTa proteomes

output_folder
├── proteomes_description
│   └── Cluster_1.tsv
│   └── Cluster_1.tsv
├── result
│   └── Cluster_1
│       └── Proteome_1.faa.gz
│       └── Proteome_2.faa.gz
│   └── Cluster_2
│       └── Proteome_3.faa.gz
│   └── Cluster_3
│       └── ...
├── tmp_proteome (can be cleaned to spare disk space using --remove-tmp option)
│   └── Proteome_1.faa.gz
│   └── Proteome_2.faa.gz
│   └── Proteome_3.faa.gz
│   └── ...
├── association_taxon_taxID.json
├── proteome_cluster_tax_id.tsv
├── esmecata_metadata_proteomes.json

The proteomes_description contains list of proteomes find by esmecata on Uniprot associated to the taxonomic annotation.

The result folder contain one sub-folder for each observation_name from the input file. Each sub-folder contains the proteome associated with the observation_name.

The tmp_proteome contains all the proteomes that have been found to be associated with one taxon.

association_taxon_taxID.json contains for each observation_name the name of the taxon and the corresponding taxon_id found with ete3.

proteome_cluster_tax_id.tsv contains the name, the taxon_id and the proteomes associated to each observation_name.

esmecata_metadata_proteomes.json is a log about the Uniprot release used and how the queries ware made (REST or SPARQL). It also gets the metadata associated to the command used with esmecata and the dependencies.

EsMeCaTa clustering

output_folder
├── cluster_founds
│   └── Cluster_1.tsv
│   └── ...
├── computed_threshold
│   └── Cluster_1.tsv
│   └── ...
├── fasta_consensus
│   └── Cluster_1.faa
│   └── ...
├── fasta_representative
│   └── Cluster_1.faa
│   └── ...
├── mmseqs_tmp (can be cleaned to spare disk space using --remove-tmp option)
│   └── Cluster_1
│       └── mmseqs intermediary files
│       └── ...
│   └── ...
├── reference_proteins
│   └── Cluster_1.tsv
│   └── ...
├── reference_proteins_consensus_fasta
│   └── Cluster_1.faa
│   └── ...
├── reference_proteins_representative_fasta
│   └── Cluster_1.faa
│   └── ...
├── proteome_cluster_tax_id.tsv
├── esmecata_metadata_clustering.json

The cluster_founds contains one tsv file per observation_name and these files contain the clustered proteins The first column contains the representative proteins of a cluster and the following columns correspond to the other proteins of the same cluster. The first protein occurs two time: one as the representative member of the cluster and a second time as a member of the cluster.

The computed_threshold folder contains the ratio of proteomes represented in a cluster compared to the total number of proteomes associated to a taxon. If the raio is equal to 1, it means that all the proteomes are representated by a protein in the cluster, 0.5 means that half of the proteoems are representated in the cluster. This score is used when giving the -t argument.

The fasta_consensus contains all the consensus proteins associated to an observation_name.

The fasta_representative contains all the representative proteins associated to an observation_name.

The mmseqs_tmp folder contains the intermediary files of mmseqs2 for each observation_name.

The reference_proteins contains one tsv file per observation_name and these files contain the clustered proteins kept after clustering process. it is similar to cluster_founds but it contains only protein kept after clsutering and threshold.

The reference_proteins_consensus_fasta contains the consensus proteins associated to an observation_name for the cluster kept after clustering process. So comapred to the fasta of fasta_consensus it is a sublist with only cluster passing the threshold.

The reference_proteins_representative_fasta contains the consensus proteins associated to an observation_name for the cluster kept after clustering process. So comapred to the fasta of fasta_representative it is a sublist with only cluster passing the threshold.

The proteome_cluster_tax_id.tsv file is the same than the one created in esmecata proteomes.

esmecata_metadata_clustering.json is a log about the the metadata associated to the command used with esmecata and the dependencies.

EsMeCaTa annotation

output_folder
├── annotation
│   └── Cluster_1.tsv
│   └── ...
├── annotation_reference
│   └── Cluster_1.tsv
│   └── ...
├── expression_annotation (if --expression option)
│   └── Cluster_1.tsv
│   └── ...
├── pathologic
│   └── Cluster_1
│       └── Cluster_1.pf
│   └── ...
│   └── taxon_id.tsv
├── uniref_annotation (if --uniref option)
│   └── Cluster_1.tsv
│   └── ...
├── esmecata_metadata_annotation.json

The annotation folder contains a tabulated file for each observation_name. It contains the annotation retrieved with Uniprot (protein_name, review, GO Terms, EC numbers, Interpros, Rhea IDs and gene name) associated to all the proteins in a proteome or associated to an observation_name.

The annotation_reference contains annotation only for the representative proteins, but the annotation of the other proteins of the same cluster can be propagated to the reference protein if the -p was used.

The expression_annotation contains expression annotation for the proteins of a taxon (if the --expression option was used).

The pathologic contains one sub-folder for each observation_name in which there is one PathoLogic file. There is also a taxon_id.tsv file which corresponds to a modified version of proteome_cluster_tax_id.tsv with only the observation_name and the taxon_id. This folder can be used as input to mpwt to reconstruct draft metabolic networks using Pathway Tools PathoLogic.

The esmecata_metadata_annotation.json serves the same purpose as the one used in esmecata proteomes to retrieve metadata about Uniprot release at the time of the query. It also gets the metadata associated to the command used with esmecata and the dependencies.

The uniref_annotation contains the annotation from the representative protein of teh UniRef cluster associated to the proteins of a taxon (if the --uniref option was used).