NTXpred is a computational platform developed for predicting neurotoxins and classifying them based on their biological source and functional activity.
The system uses multiple machine learning and sequence-analysis approaches including:
- Support Vector Machine (SVM)
- Feed Forward Neural Network (FNN)
- Recurrent Neural Network (RNN)
- PSI-BLAST
- MEME/MAST
The platform predicts:
- Neurotoxin proteins
- Source of neurotoxins
- Functional class of neurotoxins
- Ion channel blocker subclasses
Web Server:
http://www.imtech.res.in/raghava/ntxpred/
Title: Prediction of neurotoxins based on their function and source
Authors:
Sudipto Saha and Gajendra P. S. Raghava
Journal: In Silico Biology (2007)
DOI: https://doi.org/10.3233/ISB-2007-7104 https://doi.org/10.5281/zenodo.20103607
Neurotoxins are toxic proteins that affect the nervous system by blocking nerve impulses and interfering with ion channels or neurotransmitter release.
Major biological sources include:
- Eubacteria
- Cnidaria
- Mollusca
- Arthropoda
- Chordata
Neurotoxins are important for:
- Drug discovery
- Pain research
- Epilepsy therapeutics
- Ion channel studies
- Functional proteomics
The dataset was collected from Swiss-Prot/Tox-Prot databases.
- 932 experimentally validated neurotoxin proteins
- 582 neurotoxin sequences
Redundancy reduction was performed using PROSET software with 90% sequence identity cutoff.
- Eubacteria
- Cnidaria
- Mollusca
- Arthropoda
- Chordata
- Ion channel blockers
- Acetylcholine receptor blockers
- Inhibitors of acetylcholine release
- Facilitators of acetylcholine release
- Sodium channel blockers
- Potassium channel blockers
- Calcium channel blockers
- Chloride channel blockers
SVM models were developed using:
- Amino acid composition
- Dipeptide composition
- Sequence length
- Accuracy: 84.19%
- Accuracy: 92.75%
Used for similarity-based neurotoxin prediction.
Used for motif discovery and motif-based classification.
| Method | Accuracy | MCC |
|---|---|---|
| SVM Composition | 97.72% | 0.9416 |
| SVM Dipeptide | 96.05% | 0.9247 |
| RNN | 92.75% | 0.8572 |
| FNN | 84.19% | 0.6890 |
- Overall Accuracy: 92.10%
The hybrid model combined:
- PSI-BLAST
- SVM Composition + Length
- Overall Accuracy: 96.00%
The hybrid model combined:
- MEME/MAST
- SVM Dipeptide + Length
Maximum overall prediction accuracy:
- 75.08%
Classes predicted:
- Sodium channel blockers
- Potassium channel blockers
- Calcium channel blockers
- Chloride channel blockers
The study observed that neurotoxins contain:
- High cysteine content
- Conserved amino acid patterns
- Distinct amino acid composition compared to non-toxin proteins
Important enriched residues include:
- Cysteine
- Glycine
- Lysine
- Tyrosine
Hybrid approaches combined:
- SVM + PSI-BLAST
- SVM + MEME/MAST
These hybrid systems improved prediction accuracy significantly over individual methods.
- SVM_light
- SNNS Neural Networks
- PSI-BLAST
- MEME/MAST
- PROSET
- Perl
- CGI
- Linux
NTXpred can be used for:
- Neurotoxin prediction
- Venom protein analysis
- Drug discovery
- Functional proteomics
- Ion channel research
- Therapeutic peptide studies
The server allows users to:
- Predict neurotoxins
- Identify toxin source
- Predict toxin function
- Predict ion channel blocker subclasses
- Submit sequences in FASTA format
- Analyze proteins using multiple prediction methods
Web Server:
http://www.imtech.res.in/raghava/ntxpred/
Mirror Server:
http://bioinformatics.uams.edu/mirror/ntxpred
Department of Computational Biology
Indraprastha Institute of Information Technology Delhi
New Delhi, India
Email: raghava@iiitd.ac.in
Creative Commons Attribution License
Generated from the uploaded NTXpred research paper. :contentReference[oaicite:0]{index=0}