TumorHPD is a computational tool developed to predict, design, and scan tumor homing peptides. Tumor homing peptides are short peptides that specifically recognize tumor cells or tumor-associated microenvironments such as tumor vasculature after systemic delivery. This resource provides a machine learning-based approach for identifying peptides with tumor-targeting potential, which can support targeted drug delivery, tumor imaging, and peptide-based cancer therapeutics.
Web Server: http://crdd.osdd.net/raghava/tumorhpd/
Sharma, A., Kapoor, P., Gautam, A., Chaudhary, K., Kumar, R., Chauhan, J. S., Tyagi, A., and Raghava, G. P. S. Computational approach for designing tumor homing peptides. Scientific Reports, 3, 1607, 2013.
https://doi.org/10.1038/srep01607
This tool and dataset is also available on Zenodo: https://doi.org/10.5281/zenodo.20151219
Tumor homing peptides are short peptides, generally 3 to 15 amino acids long, that can specifically bind to tumor cells or tumor vasculature. These peptides are useful because they can guide therapeutic molecules, imaging agents, nanoparticles, oligonucleotides, or anticancer drugs directly to tumor sites.
Common tumor-homing motifs include RGD and NGR. RGD peptides bind to integrins, while NGR peptides bind to aminopeptidase N, which is present on tumor endothelial cells. TumorHPD was developed to computationally predict such tumor homing peptides and to help design improved peptide analogues with better tumor-targeting potential.
Data Compilation: The tool was trained on experimentally validated tumor homing peptides collected from the TumorHoPe database. Since experimentally validated non-tumor-homing peptides were not available, random peptides generated from Swiss-Prot proteins were used as negative examples.
Methodology: TumorHPD uses Support Vector Machine-based models trained on sequence-derived peptide features such as amino acid composition, dipeptide composition, and binary profile patterns.
Predictive Modeling: Allows users to submit peptide sequences and predict whether they are likely to be tumor homing peptides.
Accuracy: The best amino acid composition-based model achieved an accuracy of 86.56%, while binary profile-based models also showed strong performance with an AUC of 0.91.
Protein Scanning: Users can submit full-length protein sequences to identify possible tumor homing peptide regions.
Peptide Designing: The design module generates all possible single-substitution mutants of an input peptide and predicts whether the mutations improve tumor-homing potential.
Physicochemical Analysis: TumorHPD calculates important peptide properties such as hydrophobicity, amphipathicity, charge, and isoelectric point.
User-Friendly Interface: The server provides prediction, design, scanning, physicochemical property calculation, and secondary structure prediction in a simple web-based format.
Targeted Cancer Therapy: TumorHPD can help identify peptides that may guide therapeutic molecules specifically to tumor cells or tumor vasculature.
Drug Delivery: Predicted tumor homing peptides can be used for designing targeted delivery systems for anticancer drugs, nanoparticles, imaging agents, and oligonucleotides.
Tumor Imaging: The tool can assist in the identification of peptide candidates for tumor-specific imaging and diagnostic applications.
Peptide-Based Therapeutics: TumorHPD can help researchers design improved peptide analogues with better tumor-homing potential and useful physicochemical properties.
Cancer Bioinformatics: The tool provides a computational framework for studying sequence patterns and machine learning-based prediction of tumor-targeting peptides.
Prof. Gajendra P. S. Raghava
Corresponding Author
Email: raghava@imtech.res.in
Bioinformatics Centre
CSIR-Institute of Microbial Technology
Chandigarh-160036, India
TumorHPD was developed with support from the Council of Scientific and Industrial Research, Government of India.
The authors also acknowledge support from the Open Source Drug Discovery project, GENESIS BSC0121 project, Department of Biotechnology, Government of India, and BTISNET project.