Variational Autoencoder for the Generation of New Antimicrobial Peptides

Abstract

New techniques for antimicrobial peptide (AMP) discovery are necessary for overcoming pathogenic bacteria in the post-antibiotic era. AMPs have widely arisen via natural evolution from complex communities of competing organisms, making them promising targets for antimicrobial development; unfortunately, their identification, characterization, and production of AMPs can be complex and time consuming. This report details the development of a peptide generation framework based around variational autoencoder (VAE) and antimicrobial activity prediction models for designing novel AMPs using minimal data inputs (sequences and experimental minimum inhibitory concentration (MIC). By sampling from different, select regions of the latent space enables controlled production of new promising AMP sequences with desirable properties. Extensive analysis of the sequences and experimental validation showed this design framework as a promising system for development of novel AMPs.

Document Details

• Document Type: Technical Report
• Publication Date: Apr 08, 2021
• Accession Number: AD1127492

Entities

People

• Scott N Dean

Organizations

• United States Naval Research Laboratory

Tags

Communities of Interest

• Advanced Electronics
• Autonomy
• Materials and Manufacturing Processes

DTIC Thesaurus Topics

• Algorithms
• Anti-Bacterial Agents
• Anti-Infective Agents
• Artificial Intelligence Software
• Bacteria
• Chemical Synthesis
• Chemistry
• Computational Biology
• Computational Science
• Data Mining
• Data Science
• Databases
• Deep Learning
• Dimensionality Reduction
• Engineering
• Infection
• Information Processing
• Information Science
• Information Systems
• Machine Learning
• Microbiology
• Neural Networks
• Pathogenic Bacteria
• Predictive Modeling
• Training

Readers

• Microbial Pathology
• Neural Network Machine Learning.
• Systems Analysis and Design

Technology Areas

• Space