This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Generation of structure-guided pMHC-I libraries using Diffusion Models
Created by
Haebom
Author
Sergio Mares, Ariel Espinoza Weinberger, Nilah M. Ioannidis
Outline
To address the limitations of personalized vaccines and T-cell immunotherapies that rely on identifying peptide-MHC class I (pMHC-I) interactions capable of eliciting effective immune responses, this paper presents a structure-based benchmark of pMHC-I peptides designed based on crystal structure interaction distances using a diffusion model. This benchmark, encompassing 20 prioritized HLA alleles and independent of previously characterized peptides, demonstrates structural generalization without experimental dataset bias by reproducing canonical anchor residue notations. Using this resource, we demonstrate that state-of-the-art sequence-based predictors underperform in recognizing binding potential for these structurally stable designs, revealing allele-specific limitations not seen in existing evaluations. Our geometry-aware design pipeline generates peptides with high predicted structural integrity and higher residue diversity than existing datasets, providing a valuable resource for unbiased model training and evaluation. Code and data are available at https://github.com/sermare/struct-mhc-dev .
