TNFepitope is a host-specific computational tool developed for the prediction, design, and scanning of Tumor Necrosis Factor-alpha (TNF-α) inducing epitopes.
TNF-α is a major pro-inflammatory cytokine involved in immune regulation, inflammation, cancer, autoimmune disorders, viral infections, and cytokine release syndrome. TNFepitope helps researchers identify peptide regions that may induce TNF-α production, supporting peptide-based vaccine design, subunit vaccine development, and immunotherapy research.
TNFepitope Web Server:
https://webs.iiitd.edu.in/raghava/tnfepitope/
If you use TNFepitope in your research, please cite:
Dhall, A., Patiyal, S., Choudhury, S., Jain, S., Narang, K., and Raghava, G.P.S. TNFepitope: A webserver for the prediction of TNF-α inducing epitopes. Computers in Biology and Medicine, 160, 106929, 2023.
DOI: https://doi.org/10.1016/j.compbiomed.2023.106929
This tool and dataset is also available on Zenodo: https://doi.org/10.5281/zenodo.20150935
Tumor Necrosis Factor-alpha, also known as TNF-α, is a pleiotropic pro-inflammatory cytokine that plays an important role in:
- Immune signaling
- Inflammatory response
- Viral infection response
- Cancer biology
- Autoimmune disorders
- Cytokine release syndrome
Higher TNF-α expression has been associated with several diseases, including cancer, autoimmune disorders, rheumatoid arthritis, diabetes, Crohn’s disease, inflammatory bowel disease, HIV infection, and COVID-19-related cytokine storm.
TNFepitope was developed to identify peptide epitopes that can induce TNF-α production. These epitopes can be useful for understanding inflammatory immune responses and designing peptide-based immunotherapeutic strategies.
Experimentally validated TNF-α inducing and non-inducing epitopes were collected mainly from the Immune Epitope Database, also known as IEDB.
The study focused on two major hosts:
- Human
- Mouse
After preprocessing, peptide sequences were filtered based on length and redundancy.
| Host | TNF-α Inducing Peptides | TNF-α Non-Inducing Peptides |
|---|---|---|
| Human | 1215 | 1312 |
| Mouse | 539 | 539 |
An alternate negative dataset was also generated using randomly selected peptides from the Swiss-Prot database.
TNFepitope uses both alignment-free and alignment-based prediction approaches.
Machine learning models were trained using peptide sequence-derived features.
The classifiers used include:
- Random Forest
- Decision Tree
- Gaussian Naive Bayes
- Logistic Regression
- Support Vector Classifier
- K-Nearest Neighbor
- Extra Tree Classifier
Peptide features were generated using the Pfeature package.
Major feature types include:
- Amino Acid Composition
- Di-Peptide Composition
- Amphiphilic Pseudo Amino Acid Composition
- Atomic Composition
- Composition-Enhanced Transition Distribution
- Distance Distribution of Residues
- Pseudo Amino Acid Composition
- Physico-Chemical Properties Composition
- Quasi-Sequence Order
- Residue Repeat Information
- Conjoint Triad Descriptors
- Shannon Entropy-based features
Among these, Di-Peptide Composition-based features showed strong performance for classifying TNF-α inducing and non-inducing peptides.
TNFepitope also uses a BLAST-based similarity search approach.
In this method, a query peptide is compared against known TNF-α inducing and non-inducing peptides. The final class is assigned based on similarity with known experimentally validated epitopes.
A hybrid method was developed by combining:
- BLAST-based similarity score
- Machine learning prediction score
This hybrid approach improved the overall prediction performance.
| Host | Best Hybrid AUROC |
|---|---|
| Human | 0.83 |
| Mouse | 0.77 |
The hybrid model performed better than individual machine learning or BLAST-based approaches.
Users can submit peptide sequences to predict whether they are likely to induce TNF-α production.
Users can submit complete protein sequences to scan for potential TNF-α inducing regions.
The design module allows users to generate possible mutants of a peptide and evaluate whether the mutations increase or decrease TNF-α inducing potential.
The BLAST search module compares query peptides against a curated database of known TNF-α inducing and non-inducing epitopes.
A standalone version is available for users who want to run predictions locally.
| Module | Description |
|---|---|
| Predict | Predict whether a peptide is TNF-α inducing or non-inducing |
| Design | Generate peptide mutants and predict their TNF-α inducing potential |
| Scan | Scan full-length proteins for TNF-α inducing regions |
| BLAST Search | Search query peptides against known TNF-α inducing and non-inducing epitopes |
| Standalone | Run TNFepitope locally |
TNFepitope can help identify peptide regions that may induce TNF-α-mediated immune responses and may be considered during subunit vaccine development.
The tool can be used to scan viral proteins for TNF-α inducing regions, which may help in understanding inflammatory responses during viral infections.
Since TNF-α is involved in autoimmune and inflammatory diseases, TNFepitope may help identify peptides associated with immune activation.
The design module can help researchers modify peptide sequences and study how mutations affect TNF-α inducing potential.
TNFepitope provides a computational framework for studying cytokine-inducing epitopes using sequence-based machine learning and similarity-based methods.
Prof. Gajendra P. S. Raghava
Department of Computational Biology
Indraprastha Institute of Information Technology Delhi
Okhla Phase III, New Delhi, India
Email: raghava@iiitd.ac.in
Lab Website: https://webs.iiitd.edu.in/raghava/
This work received support from the Department of Biotechnology, Government of India.
The authors also acknowledge support from the Department of Science and Technology INSPIRE program and infrastructure provided by the Department of Computational Biology, IIIT Delhi.
This resource is available for academic and research use.
Users are requested to cite the original publication if they use TNFepitope in their work.
The authors acknowledge the Department of Biotechnology, Government of India, the Department of Science and Technology INSPIRE program, and the Department of Computational Biology, IIIT Delhi, for financial support, fellowships, infrastructure, and facilities used in developing TNFepitope.