Anton Bushuiev

Despite major advances in computational protein design, reliably engineering new proteins and their interactions for targets not seen during training remains difficult. The goal of this PhD project is to develop machine learning methods that generalize to novel proteins by learning effectively from 3D structural data, improving the quality of training datasets and evaluation benchmarks, and adapting models to a specific target protein at inference time. The project spans both de novo binder design and protein engineering via mutation. The resulting approaches are validated on practical protein engineering tasks, including applications with therapeutic potential, where strong generalization is particularly important because real-world targets are often novel or not well studied.