Researchers have developed a novel attack called the "Genotypic Trigger" that can intentionally induce health risks in generative antimicrobial peptide (AMP) models. This attack manipulates models to produce peptides with a significantly elevated predicted immunogenicity risk for individuals carrying specific HLA alleles, while posing little risk to others. Despite this targeted risk, the backdoored models maintain their primary functions, such as antimicrobial potency and low general toxicity, allowing them to bypass standard safety screenings. AI
IMPACT This research highlights potential security vulnerabilities in AI models used for drug discovery, necessitating new safety protocols for generative AI in sensitive applications.
RANK_REASON The cluster contains a research paper detailing a novel attack method on AI models used in drug discovery. [lever_c_demoted from research: ic=1 ai=1.0]
- Antimicrobial peptides
- arXiv
- Doniyorkhon Obidov
- Genotypic Trigger
- HLA allele
- large-language models
- peptide generation models
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
