Related Experiment Video
Updated: Aug 14, 2025
A Combined 3D Tissue Engineered In Vitro/In Silico Lung Tumor Model for Predicting Drug Effectiveness in Specific Mutational Backgrounds
Published on: April 6, 2016
LGBM-ACp: an ensemble model for anticancer peptide prediction and in silico screening with potential drug targets
Swarnava Garai1, Juanit Thomas1, Palash Dey2
1Department of Bioengineering, NIT Agartala, Tripura, 799046, India.
Abstract:
Conventional cancer therapies are highly expensive and have serious complications. An alternative approach now emphasizes on the development of small, biologically active peptides without acute toxicity. Experimental screening to find curative anticancer peptides (ACP) often gives rise to multiple obstacles and is time dependent. Consequently, developing an effective computational technique to identify promising ACP candidates prior to preclinical research is in high demand. This study proposed a machine-learning framework that used the light gradient-boosting machine as a classifier and two compositional and two binary profile features as input. The ensemble model displayed an accuracy, MCC, and AUROC of 97.52%, 0.91, and 0.98, respectively, which outclassed most of the existing sequence-based computational tools. A distinct dataset of non-mutagenic, non-toxic, and non-inhibitory Cytochrome P-450 peptides was used to validate the hybrid model. The most relevant ACP in the alternative dataset was compared with two standard ACPs, beta defensin 2, and cecropin-A. Molecular docking of the predicted peptide revealed that it has a strong binding affinity with twenty-five anticancer drug targets, most notably phosphoenolpyruvate carboxykinase (- 7.2 kcal/mol). Additionally, molecular dynamics simulation and principal component analysis supported the stability of the peptide-receptor complex. Overall, the present findings will take a step forward in rational drug design through rapid identification and screening of therapeutic peptides.
Insights
This study introduces a machine learning model to rapidly identify effective anticancer peptides (ACPs). The developed computational framework offers a promising alternative to traditional therapies, accelerating drug discovery.
Area of Science:
- Biotechnology
- Computational Biology
- Drug Discovery
Background:
- Conventional cancer therapies face challenges including high costs and severe side effects.
- There is a growing need for novel, less toxic therapeutic agents like anticancer peptides (ACPs).
- Experimental screening for ACPs is often time-consuming and resource-intensive.
Purpose of the Study:
- To develop an efficient computational method for identifying promising ACP candidates.
- To overcome the limitations of traditional experimental screening methods for ACPs.
- To accelerate the preclinical research and development of novel peptide-based cancer therapies.
Main Methods:
- A machine learning framework utilizing a light gradient-boosting machine classifier.
- Integration of compositional and binary profile features for peptide analysis.
- Validation using a distinct dataset of non-mutagenic, non-toxic peptides and comparison with standard ACPs.
- Molecular docking and dynamics simulations to assess binding affinity and complex stability.
Main Results:
- The ensemble model achieved high performance metrics: 97.52% accuracy, 0.91 MCC, and 0.98 AUROC.
- The model outperformed existing sequence-based computational tools for ACP identification.
- The predicted peptide demonstrated strong binding affinity to multiple anticancer drug targets, including phosphoenolpyruvate carboxykinase.
- Molecular simulations confirmed the stability of the peptide-receptor complex.
Conclusions:
- The proposed machine learning framework significantly enhances the rapid identification and screening of potential anticancer peptides.
- This computational approach represents a significant advancement in rational drug design for peptide-based cancer therapeutics.
- The findings pave the way for more efficient development of novel and effective cancer treatments.
