HMAMP: Hypervolume-Driven Multi-Objective Antimicrobial Peptides Design

TL;DR

This study tackles the challenge of simultaneously optimizing multiple properties of antimicrobial peptides (AMPs) by introducing HMAMP, a Hypervolume-driven Multi-objective AMP design framework. HMAMP combines a multi-objective GAN with reinforcement learning and hypervolume-based gradient descent to expand exploration and stabilize training, while predicting MIC and hemolysis/toxicity via fine-tuned Prot-BERT-BFD predictors to construct a Pareto front and identify knee points with the Kneedle algorithm. The approach yields 5000 candidate AMPs and analyzes top knee-point solutions through structural predictions (Alphafold) and secondary-structure inference, confirming alpha-helix rich, cationic, amphiphilic characteristics. Across multiple benchmarks, HMAMP achieves higher hypervolume and greater Pareto coverage than state-of-the-art baselines, demonstrating robust multi-objective optimization, diversity, and practical potential for multi-attribute AMP design.

Abstract

Antimicrobial peptides (AMPs) have exhibited unprecedented potential as biomaterials in combating multidrug-resistant bacteria. Despite the increasing adoption of artificial intelligence for novel AMP design, challenges pertaining to conflicting attributes such as activity, hemolysis, and toxicity have significantly impeded the progress of researchers. This paper introduces a paradigm shift by considering multiple attributes in AMP design. Presented herein is a novel approach termed Hypervolume-driven Multi-objective Antimicrobial Peptide Design (HMAMP), which prioritizes the simultaneous optimization of multiple attributes of AMPs. By synergizing reinforcement learning and a gradient descent algorithm rooted in the hypervolume maximization concept, HMAMP effectively expands exploration space and mitigates the issue of pattern collapse. This method generates a wide array of prospective AMP candidates that strike a balance among diverse attributes. Furthermore, we pinpoint knee points along the Pareto front of these candidate AMPs. Empirical results across five benchmark models substantiate that HMAMP-designed AMPs exhibit competitive performance and heightened diversity. A detailed analysis of the helical structures and molecular dynamics simulations for ten potential candidate AMPs validates the superiority of HMAMP in the realm of multi-objective AMP design. The ability of HMAMP to systematically craft AMPs considering multiple attributes marks a pioneering milestone, establishing a universal computational framework for the multi-objective design of AMPs.

Figures (7)

• Figure 1: Antimicrobial peptides(AMPs) cross cell membranes and act on bacteria and red blood cells, and low MIC and hemolysis conflict.
• Figure 2: The challenges we are attempting to address can be summarized as follows: Firstly, AMPs are composed of 20 different amino acids with a sequence length of less than 100, resulting in an enormous sequence space. Secondly, there are apparent conflicts in the properties of AMPs, and our objective is to simultaneously minimize multiple attributes (such as MIC, and hemolysis) to generate a balanced Pareto-optimal set considering various properties of AMPs. Finally, it is crucial to identify solutions on the optimal Pareto frontier that are of particular interest to decision-makers, such as the most promising knee points.
• Figure 3: Overview of the proposed HMAMP (shown here as a two-discriminator setup). Block(A) depicts the dataset of AMPs used for pretraining the generative model, with two additional target datasets contributing to the training of discriminators and predictors. Block(B) illustrates the application of the hypervolume maximization concept for training a multiple-discriminator generative model. Block(C) demonstrates the capabilities of the fine-tuned classifier and discriminator, achieved through Pro-BERT-BFD-based fine-tuning, for predicting the generated candidate AMPs. Block(D) displays the Pareto front of the generated AMPs along with the most promising knee point solutions, representing optimal AMP candidates. Block(E) portrays the AMPs module, subject to screening through helical structure analysis and molecular dynamics simulation, to validate the physicochemical attributes and secondary structure of the candidate AMPs.
• Figure 4: 2D example of the objective space where the generator loss is being optimized.
• Figure 5: The t-SNE plot illustrating the distribution of amino acids in each sequence generated by HMAMP and three datasets ($D_{\text{AMP}}$, $D_{\text{MIC}}$, and $D_{\text{HEMO}}$) .