Peptide-Based Drug Predictions for Cancer Therapy Using Deep Learning.

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Anticancer peptides (ACPs) show promise as targeted cancer therapeutics.
• Traditional methods for identifying ACPs are costly and time-consuming.
• There is a need for efficient in silico methods to screen potential ACPs.

Purpose of the Study:

• To develop a novel computational method for predicting anticancer peptide (ACP) activity.
• To create a user-friendly web-based tool for early-stage evaluation of potential ACPs.
• To accelerate the discovery and development of new anticancer peptide drugs.

Main Methods:

• Collected and curated an updated dataset of known anticancer peptides (ACPs).
• Developed PC6, a novel peptide sequence encoding method representing six physicochemical properties.
• Trained a convolutional neural network (CNN) deep learning model using the PC6 encoding method and ACP dataset.
• Created AI4ACP, a web-based platform integrating the dataset, encoding method, and CNN model.

Main Results:

• The AI4ACP tool demonstrated superior performance compared to existing ACP predictors.
• 5-fold cross-validation showed stable high accuracy (around 0.89) without overfitting.
• The PC6 encoding method effectively represents peptide sequences for deep learning models.
• AI4ACP enables rapid in silico prediction of anticancer properties for novel peptides.

Conclusions:

• AI4ACP provides an efficient and accurate platform for the initial screening of anticancer peptides.
• The developed method accelerates the drug discovery pipeline for novel ACPs.
• AI4ACP facilitates the selection of promising peptide candidates for experimental validation.