The project was initially developed in Perl, in 2006, for Linux-users only. You can find more info about it in the related article, clicking on this link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1570458
Now, we are carrying on this project independently from the original developer and this is the Github page of the NERVE 2.0-stand-alone version.
Users can also leverage a user-friendly web-based graphical interface (GUI) hosted on our dedicated server: https://nerve-bio.org. This GUI empowers even novice users to operate NERVE with ease, requiring only the proteome FASTA file as input.
It further enhances the experience by allowing visualization of analysis results using our custom-built visualizer, also facilitating the identification of epitopes (and their constituent parts) predicted to bind to multiple HLA alleles.
- Machine Learning and Alignment Analysis, using ML methods for antigen analysis
- Flexible Usage, customizing your settings
- Stand-alone Docker Version, reducing dependencies for its installation
- Intuitive Web-based GUI,
- Complete Python Rewrite, enhancing its performance and maintainability.
- Ongoing Development, continuously evolving with new features and improvements.
- Comprehensive Output, including:
- a CSV file summarizing all candidate proteins with scores and predicted characteristics (P_adhesin, P_antigen, location, etc.).
- a separate CSV file listing discarded proteins.
- dedicated folders containing additional files (CSVs and PNGs) for each selected protein with predicted linear epitopes.
Benefits for Vaccine Research:
- Accessibility: researchers worldwide can access and use NERVE without barriers, accelerating vaccine development efforts globally.
- Transparency: allowing researchers to scrutinize the code, identify potential issues, and contribute to its refinement.
- Collaboration: by enabling researchers to share, modify, and extend the pipeline, leading to innovative advancements.
- Community-Driven Development: the open-source model encourages community participation, ensuring that NERVE remains aligned with the evolving needs of the vaccine research community.
Whether it's reporting bugs, suggesting improvements, or developing new features, your input is valuable.
We welcome contributions from researchers interested in shaping the future of NERVE!
NERVE can be used as a stand alone version taking advantage of Docker in Linux systems.
1) Install Docker following these instructions.
2) Follow this post-installation procedure even if you already installed Docker.
2) Clone the repository:
git clone https://github.com/nerve-bio/NERVE.git
3) Navigate to the correct folder:
cd NERVE
4) Run NERVE (the first time it will take a few minutes)
./NERVE.sh --help
This is the expected output:
usage: NERVE.py [-h] [-a] [-ev] -g [-ml] [-mm] [-m] [-mpsl] -p1 [-p2] [-pl] [-rz] [-rl] [-s] [-ss] [-tdl] [-vl] [-vir]
[-ep] [-m1l] [-m2l] [-m1ovr] [-m2ovr] [-prt] [-wd] [-nd] [-id] [-dfd] [-epp]
NERVE arguments:
-h, --help show this help message and exit
-a, --annotation Activation (True) or deactivation (False) of annotation module. Uses DeepFri to retrieve protein functional onthologies (default: True)
-ev, --e_value Expect-value used in blastp for immunity modules (default: 1e-10)
-g, --gram Negative (n) or positive (p) gram stain of the pathogen of interest (default: None)
-ml, --minlength Minimal length required for shared peptides to be extracted in comparison analyses versus human and/or mouse (default: 9)
-mm, --mismatch Maximal number of not compatible substitutions allowed in shared peptides alignment windows of 'minlength' size in immunity modules (default: 1)
-m, --mouse Activation (True) or deactivation (False) of the mouse immunity module. This module compares proteome1 with mouse proteome and a further analysis of the
eventual shared peptides is carried out as in the autoimmunity module (default: False)
-mpsl, --mouse_peptides_sum_limit
Parameter calculated in mouse module and used by select module. Protein with 'sum of shared peptides of the i-protein with mouse proteins/number of aminoacids
of the i-protein' <= mouse_peptides_sum_limit and with absence of match mhc-I and Mhc-II mouse ligands are selected (default: 0.15)
-p1, --proteome1 Path to proteome or Uniprot proteome ID (see: https://www.uniprot.org/proteomes/?query=&sort=score) (default: None)
-p2, --proteome2 Path to proteome or Uniprot proteome ID (see: https://www.uniprot.org/proteomes/?query=&sort=score) (default: None)
-pl, --padlimit Set the probability of adhesin (pad) value cut-off for all proteins in select module. Thus, these proteins with a pad value < cut-
off are discarded (0.-1) (default: 0.5)
-rz, --razor Activation (True) or deactivation (False) of the loop-razor module. This module allows the recovery of protein vaccine candidates, with more than 2
transmembrane domains, that would otherwise be discarded in the select module. The longest loop with minimum len == 'razlen' aa will replace the original
protein sequence for following NERVE steps (default: False)
-rl, --razlen Set minimal length of loop considered in loop-razor module (default: 50)
-s, --select Activation (True) or deactivation (False) of select module, which filters PVC from proteome1 (default: True)
-ss, --substitution
Maximal number of compatible substitutions allowed in shared peptides alignment windows of 'minlength' size in immunity modules (default: 3)
-tdl, --transmemb_doms_limit
Parameter of select module. Proteins with trasmembrane domains >= transmemb_doms_limit are discarded (default: 3)
-vl, --virlimit Cut-off value for NERVirulent in the select module (0.-1) (default: 0.5)
-vir, --virulent Activation (True) or deactivation (False) of NERVirulent module, predictor of the probability of being a virulence factor (default: False)
-ep, --epitopes Activate or deactivate epitope prediction module (default: True)
-m1l, --mhci_length
mhci binders length (9, 10, 11 are available) (default: 9)
-m2l, --mhcii_length
mhcii binders length (9, 11, 12, 15 are available) (default: 11)
-m1ovr, --mhci_overlap
mhci-epitope overlap (default: 1)
-m2ovr, --mhcii_overlap
mhcii-epitope overlap (default: 1)
-prt, --epitope_percentile
percentile decision threshold on which to predict epitopes from full length proteins (default: 0.9)
-wd, --working_dir
Path to working directory. If not existing, a working directory with the given path is created (default: ./)
-nd, --NERVE_dir Path to NERVE repository folder (download from: https://github.com/nicolagulmini/NERVE) (default: /usr/nerve_python/NERVE)
-id, --iFeature_dir
Path to iFeature repository folder (download from: https://github.com/Superzchen/iFeature) (default: /usr/nerve_python/assets/iFeature)
-dfd, --DeepFri_dir
Path to DeepFri folder (download from: https://github.com/flatironinstitute/DeepFRI) (default: /usr/nerve_python/assets/DeepFri)
-epp, --ep_plots Epitopes plotting script (default: True)
5) (optional) test correct functioning:
pytest tests/tests.py
With help, you can visualize all possible arguments, which include all user-settable parameters and activation or deactivation of some NERVE components. Definitions and setting options are shown in detail for each of them here above. Important: always specify relative paths for input files and working directory.
Mandatory arguments:
-p1,--proteome1-g,--gram
All the other showed arguments are optional. So, if they're not set by the user, their related default value is used during NERVE computation.
To start with NERVE, the -p1,--proteome1 file, which contains all bacterial FASTA proteins to be analyzed, has to be put in the NERVE folder.
./NERVE.sh -p1 anthracis.fasta -g p
Here, the example file "anthracis.fasta", with Gram positive proteins, is saved in the NERVE folder.
But the user can also retrieve a specific proteome from Uniprot, writing its correct ID:
./NERVE.sh -p1 UP000000594 -g p
Here, the reported Uniprot proteome ID is the one of Bacillus anthracis (strain Ames Ancestor).
The next step is the activation/deactivation of facultative components and their parameters settings.
Facultative components are: Annotation, Loop-Razor, Mouse immunity, NERVirulent, Conservation, Select and Epitope prediction. Only the last two ones are acttivated by default.
See last section for setting examples.This is an output example, showing the percentage of analysis completion:
./NERVE.sh -p1 test.fasta -g n
wrote config file /root/.config/epitopepredict/default.conf
Start NERVE 2.0
10% done
20% done
30% done
40% done
50% done
60% done
70% done
80% done
90% done
100% done
End NERVE computation successfully.
The default working directory, where all result files produced during NERVE analysis are saved, is the NERVE folder.
To change it, use -wd,--working_dir as showed:
./NERVE.sh -p1 UP000000594 -g p -wd ./chosen_path
Here are some different examples to show how to correctly set NERVE arguments:
./NERVE.sh -p1 test.fasta -g n
- This is a run example with the minimum number of arguments (the two mandatory ones). Here we're considering a FASTA file (test.fasta) with Gram negative proteins (
-g n).
./NERVE.sh -p1 test.fasta -g n -s False -vir True -vl 0.8
- The same file with Gram negative proteins will be analyzed activating NERVirulent and setting its cut-off to 0.8 (instead of default " 0.5 "). Without activating Select (
-s False), PVCs are not filtered. So, in the results file, there will be the list of all proteins with their related analyzed info.
./NERVE.sh -p1 test.fasta -g n -m True -rz True -rl 75
- Here, two facultative components are activated (Mouse immunity and Loop-razor).Then, minimum length of considered loop, with Loop-razor, is set to 75aa (from 50aa).
./NERVE.sh -p1 test.fasta -g n -p2 test2.fasta
- In this case, adding test2.fasta, the component Conservation is automatically activated, allowing the user to infer antigen conservation.
For more info about components working visit the FAQ section of https://nerve-bio.org