Published April 29, 2026 | Version v1.0

ToxinPred2: an improved method for predicting toxicity of proteins

  • 1. ROR icon Indraprastha Institute of Information Technology Delhi

Description

Title:
ToxinPred2 Dataset – Experimentally validated toxic and non-toxic proteins (non-redundant, multiple datasets)

Description:

Project: ToxinPred2 – An improved method for predicting toxicity of proteins

Publication: Sharma, N., Naorem, L.D., Jain, S., & Raghava, G.P.S. (2022). ToxinPred2: an improved method for predicting toxicity of proteins. Briefings in Bioinformatics, 23(5), bbac174. https://doi.org/10.1093/bib/bbac174

Overview: This dataset accompanies ToxinPred2, an upgraded version of ToxinPred designed for predicting toxicity of proteins (ToxinPred was limited to peptides ≤35 aa). It provides three datasets curated from UniProt/Swiss-Prot for training and evaluating toxicity predictors.

Content:

Dataset Toxic proteins Non‑toxic proteins Redundancy (CD‑HIT)
Main 8,233 8,233 No (raw)
Alternate 1,924 1,924 Yes (≤40% identity)
Realistic 1,924 19,240 Yes (1:10 ratio)

 

Best Model Performance (RF + BLAST + MERCI – hybrid, main dataset):

Metric Training Validation
AUC 0.98 0.99
MCC 0.91 0.91
Accuracy 95.3% 95.5%
Sensitivity 92.9% 93.7%
Specificity 97.6% 97.4%

Comparison with existing methods (validation dataset of ToxinPred2):

Method AUC Notes
ToxinPred2 0.99 Proteins
TOXIFY 0.88 Limited to length ≤500 aa

Usage: Predicting toxicity of therapeutic proteins, screening protein libraries, complementing ToxinPred (for peptides).

Related Resources: Web server: https://webs.iiitd.edu.in/raghava/toxinpred2/ | GitHub: https://github.com/raghavagps/toxinpred2

Contact: raghava@iiitd.ac.in (Gajendra P. S. Raghava)

Files

raghavagps/toxinpred2-v1.0.zip

Files (13.2 MB)

Name Size Download all
md5:506608130d5312045f210622db27ab9e
13.2 MB Preview Download

Additional details

Related works