CancerGram: An Effective Classifier for Differentiating Anticancer from Antimicrobial Peptides

Area of Science:

• Biochemistry
• Computational Biology
• Oncology

Background:

• Antimicrobial peptides (AMPs) protect organisms and compete with microbes.
• Some AMPs, termed anticancer peptides (ACPs), target cancer cells.
• ACPs and AMPs interact with cell membranes due to their physicochemical properties.

Purpose of the Study:

• To identify novel cationic peptides that selectively eliminate cancer cells.
• To develop a computational tool for predicting anticancer peptides (ACPs) to reduce experimental costs.
• To create a robust classifier for distinguishing ACPs from other peptide types.

Main Methods:

• Utilized n-grams and random forests for peptide classification.
• Developed CancerGram, a computational tool for predicting ACPs.
• Implemented a three-class classification model: ACPs, AMPs, and non-ACPs/non-AMPs.

Main Results:

• CancerGram achieved an AU1U of 0.89 and a Kappa statistic of 0.65.
• The tool effectively classifies peptides into three distinct categories.
• Demonstrated superior performance compared to existing ACP classifiers.

Conclusions:

• CancerGram is a valuable computational tool for identifying potential anticancer peptides.
• The tool aids in the discovery of new cancer therapeutics.
• CancerGram is accessible as a web server and R package, facilitating research.