Gene-gene interaction screening pipeline

scbirlab/nf-ggi is a Nextflow pipeline to screen gene-gene interactions within an organism or between organisms (in the case of host-pathogen or phage-bacterium interactions).

Table of contents

• Processing steps
• Requirements
• Quick start
• Inputs
• Outputs
• Credit
• Issues, problems, suggestions
• Further help

Processing steps

scbirlab/nf-ggi carries out the following steps:

1. [Optional] Download Rhea DB (of metabolites) in preparation for searching.

For proteins or proteomes in the sample sheet:

1. [Optional] Download its STRING database and tidy up the data.
2. Download FASTA sequences of proteins from UniProt
   • If multiple proteomes are available, choose according to this priority: "Reference and representative", "Reference", "Representative", "Other"
3. Find reactions in Rhea DB and connect products with reactants between enzymes in the proteome.

For each FASTA sequence:

4. Generate a multiple sequence alignment with hhblits.

For method == "self":

5. Within each organism, generate all unique pairs of proteins.

For method == "bait":

5. All unique pairs of proteins between the organism and listed baits.

For method == "custom":

5. All unique pairs of proteins listed.

Then for each protein pair, optionally:

6. with --dca: Calculate the co-evolutionary signal with DCA, optionally generating plots of contact maps.
7. with --rf2t: Predict the interface contact map with yunta rf2t (RosettaFold-2track), optionally generating plots of contact maps.
8. with --af2: Predict the protein-protein complex structure map with yunta af2 (AlphaFold2), optionally generating plots of contact maps.

Requirements

You need access to the UniClust and BFD databases, and you need Nextflow and either conda, Singularity, or Docker to be installed.

Databases

To generate multiple-sequence alignments (MSAs) for co-evolutionary analysis, hhblits databases of pre-clustered sequences is required. Unfortunately, these are extremely large, so cannot be downlaoded as part of the pipeline. You should download the UniClust and BFD databases, then set the --uniclust and --bfd parameters of the pipeline (see below).

If you're at the Crick, these databases already reside on NEMO, and there is no need to downlaod them.

Software

You need to have Nextflow and either Singularity, Docker, of conda installed on your system.

First time using Nextflow?

Crick users

If you're at the Crick or your shared cluster has Nextflow and Singularity already installed, try:

module load Nextflow Singularity

Others

Otherwise, if it's your first time using Nextflow on your system, you can install it using conda:

conda install -c bioconda nextflow 

You may need to set the NXF_HOME environment variable. For example,

mkdir -p ~/.nextflow
export NXF_HOME=~/.nextflow

To make this a permanent change, you can do something like the following:

mkdir -p ~/.nextflow
echo "export NXF_HOME=~/.nextflow" >> ~/.bash_profile
source ~/.bash_profile

Quick start

There are three run modes for the pipeline:

• "self": run all protein-protein interactions within an organism
• "bait": run interactions between all proteins from an organism and either one protein or another organism's proteome
• "custom": run specified protein pairs from a file

The easiest way to get going is by specifying parameters on the command-line:

bfd=path/to/your/bfd
uniclust=path/to/your/uniclust
nextflow run scbirlab/nf-ggi \
    --bfd "$bfd" --uniclust "$uniclust" \
    --organism_id 243273 \
    --dca --rf2t  --plots

Here's what the flags mean:

• --organism_id: The Taxon ID of the organism, whih you can find at NCBI or UniProt
• --dca, --rf2t: Run direct-coupling analysis and RosettaFold-2track
• --plots: Generate amino acid contact maps. This takes about 10MB per protein-protein interaction, so be sure you have enough disk space for the number of protein-protein pairs you're testing!

You can also run --metabolites to get the metabolic network, and --string to get the STRING co-expression network.

Because only --organism_id was provided, the pipeline assumes "self" mode (i.e. all-vs-all within taxon 243273).

You can run bait mode by providing --bait <UniProt ID>:

nextflow run scbirlab/nf-ggi --bfd "$bfd" --uniclust "$uniclust" \
    --organism_id 559292 --bait P00931 \
    --dca

The bait can be another organisms's proteome. In this case, we need to specifiy that --bait_is_taxon and --interspecies:

nextflow run scbirlab/nf-ggi --bfd "$bfd" --uniclust "$uniclust" \
    --organism_id 559292 --bait 1773 \
    --bait_is_taxon --interspecies \
    --rf2t

Running with Singularity, Docker, or Conda

scbirlab/nf-ggi runs on a Singularity container engine by default to ensure software versions are consistent. If you have docker installed, you can run using -with-docker to use it instead, or if you have Conda you can run -with-conda.

Running more than one query in parallel

Make a sample sheet (see below) with columns representing the flags above, and, optionally, a nextflow.config file in the directory where you want the pipeline to run. Then simply run:

nextflow run scbirlab/nf-ggi

Pipeline versions

Each time you run the pipeline after the first time, Nextflow will use a locally-cached version which will not be automatically updated. If you want to ensure that you're using the very latest version of the pipeline, use the -latest flag.

nextflow run scbirlab/nf-ggi -latest

If you want to run a particular tagged version of the pipeline, such as v0.0.5, you can do so using

nextflow run scbirlab/nf-ggi -r v0.0.5

For help, use nextflow run scbirlab/nf-ggi --help.

The first time you run the pipeline on your system, the software dependencies in environment.yml will be installed. This may take several minutes.

Inputs

Command-line usage

The pipeline can be run with command-line arguments:

# intra-species all-vs-all:
nextflow run scbirlab/nf-ggi --uniclust <path> --bfd <path> \
    --organism_id <taxon ID>
# intra-species all-vs-1:
nextflow run scbirlab/nf-ggi --uniclust <path> --bfd <path> \
    --organism_id <taxon ID> --bait <UniProtID>
# inter-species all-vs-all:
nextflow run scbirlab/nf-ggi --uniclust <path> --bfd <path> \
    --organism_id <taxon ID> --bait <taxon ID> \
    --bait_is_taxon --interspecies
# custom list of pairs:
nextflow run scbirlab/nf-ggi --uniclust <path> --bfd <path> \
    --organism_id <taxon ID> \
    --filename <path> --column1 <gene-col1> --column2 <gene-col2> \
    [--interspecies --organism_id2 <taxon ID>] [--format <gene-name-type>]

The following parameters are required:

--organism_id             Taxon ID for organism
Bait mode:
    --bait                 UniProt ID for bait protein, or Taxon ID for bait organism
Custom mode:
    --filename             Filename to get custom protein pairs
    --column1, --column2   Column names from --filename to get protein IDs

The following parameters are optional. They have default values which can be overridden if necessary.

--reviewed      Only pull SwissProt reviewed proteins from proteome
--isoforms       Additionally pull isoform sequences from proteome
--proteome_opts  Additonal filters for pulling from proteome. Check 
           https://www.ebi.ac.uk/proteins/api/doc/#!/proteins/search for options.
--bait_is_taxon  Indicate that bait is an organism ID
--interspecies   Run analysis between interacting species proteomes
--plots          Generate contact map plots
--organism_id2   When providing a file of pairs, if the second protein (--column2) is from another organism than the first
--format         Type of gene identifier in --column1, --column2. Default: "Gene_Name"
--test           Whether to run in test mode. Default: false.
--outputs        Output folder. Default: "outputs".
--batch_size     What size to batch protein-protein interactions into. Default: 100.

Sample sheet

You can run multiple combinations in one command using a sample sheet. The sample sheet is a CSV file with one row per combination of parameters to run. The column headings have the same names as the required flags for command-line usage. The optional flags are still on the command line, and applied to everything in the run. Here, the mode needs to be specified with --mode:

nextflow run scbirlab/nf-ggi --mode self --rf2t --metabolites --sample_sheet path/to/sample-sheet.csv

Sample sheet structure

Here is an example of the sample sheet for mode = "self", to find all the mycoplasma protein-protein interactions:

organism_id	proteome_name
243273	"Mycoplasma genitalium"

The proteome_name column is not neccesary, but you can add extra columns with human-readable annotations for your own sanity. We recommend:

• proteome_name
• (if using a bait) bait_name

If running with mode = "bait", to do a pulldown against a single bait protein, add another column with the bait UniProt ID.

organism_id	proteome_name	bait	bait_name
243273	Mycoplasma genitalium	P47259	FolD

If running with mode = "custom", to do a pulldown against a single bait protein, add another column with the bait UniProt ID.

organism_id	proteome_name	format	filename	column1	column2
559292	Saccharomyces cerevisiae	Gene_Name	combos.csv	query_orf	array_orf

In this case, combos.csv must be in the inputs folder defined above. It would look like:

query_orf	query_gene_name	array_orf	array_gene_name
YAL058W	CNE1	YAL068C	PAU8

Further examples are in the test directory of this repository.

Config-file usage (recommended)

For reproducibility, self-documentation, and to save typing, parameters with the same names as the command line flags above can be provided in a nextflow.config file in the working directory. For example:

params {
    organism_id = "243273"
    dca = true
    rf2t = true
}

Or with a sample sheet:

params {
    sample_sheet = "path/to/sample-sheet.csv"
    mode = "self"
    dca = true
    rf2t = true
    metabolites = true
    string = true
}

Outputs

Outputs are saved in the output folder defined above. They include these directories:

• string: STRING co-expression values
• metabolites: Reconstructed metabolic network
• msa: All MSA files
• ppi: All protein-protein interaction data
• sequences: Protein sequences

Credit

The idea of using DCA, RoseTTAFold-2track, and AlphaFold2 in a cascade of increasingly expensive and specific PPI detection methods has been explored in a series of papers from David Baker's lab:

• Cong et al., Protein interaction networks revealed by proteome coevolution. Science, 2019
• Humpreys et al., Computed structures of core eukaryotic protein complexes. Science, 2021
• Humpreys et al., Protein interactions in human pathogens revealed through deep learning. Nature Microbiology, 2024

scbirlab/nf-ggi applies these algorithms in a Nextflow pipeline to allow easy scaling, and enables inter-species interactions. It also reconstructs metabolic networks, and pulls known interactions from the STRING database.

Issues, problems, suggestions

Add to the issue tracker.

Further help

Here are the pages of the software and databases used by this pipeline.

Databases:

• STRING for co-expression
• Rhea for enzyme reactions
• UniProt for protein sequences
• NCBI Genbank for taxonomy

Software:

• hhblits for generating MSAs
• rdkit for cheminformatics of enzyme reactants and products
• yunta for running DCA, RosettaFold-2track, and AlphaFold2 on MSAs