OmegAMP: Targeted AMP Discovery via Biologically Informed Generation

Diogo Soares, Leon Hetzel, Paulina Szymczak, Marcelo Der Torossian Torres, Johanna Sommer, Cesar de la Fuente-Nunez, Fabian J Theis, Stephan Günnemann, Ewa Szczurek

Keywords: antimicrobial peptides (AMPs), controllable generation, biologically-informed representation

TL;DR: OmegAMP improves AMP discovery by providing controlled, targeted generation enabled by versatile conditioning & bio-informed encoding. It achieves an unprecedented 96% success rate in wet-lab validation against multi-drug resistant microbes.

Abstract: Antimicrobial peptide (AMP) discovery is often limited by poor controllability and low experimental hit rates. We introduce OmegAMP, a diffusion-based framework for reliable AMP generation with fine-grained control over physicochemical properties and activity profiles. OmegAMP leverages a biologically informed encoding space and a novel synthetic data augmentation strategy for classifier-based filtering, which significantly reduces false positive rates. *In silico* experiments demonstrate state-of-the-art performance across the discovery pipeline. Crucially, in wet-lab validation, 24 out of 25 (96\%) OmegAMP-designed peptides demonstrated antimicrobial activity, including effectiveness against multi-drug resistant strains. Our results highlight OmegAMP's potential to accelerate the development of novel therapeutics against antimicrobial resistance.

Presenter: ~Diogo_Soares1

Format: Maybe: the presenting author will attend in person, contingent on other factors that still need to be determined (e.g., visa, funding).

Funding: Yes, the presenting author of this submission falls under ICLR’s funding aims, and funding would significantly impact their ability to attend the workshop in person.

Submission Number: 46