Ruxandra I. Dima

This study investigates augmentation strategies to improve protein language model performance by reducing dataset size and enhancing generalization through active learning and generative models.

The paper introduces FlexENN, a graph neural network that predicts binding energies for both folded and intrinsically disordered protein complexes, showing improved performance over rigid-body methods.

This study investigates allosteric regulation mechanisms in AAA+ ATPases using katanin and ClpB, revealing similar and specific mechanisms through simulations and machine learning.

This study investigates how different tubulin CTT isotypes influence katanin structure and dynamics, revealing that inhibitory CTTs enhance pore narrowing and flexibility, while non-inhibitory CTTs disrupt coordination.

The paper investigates how hereditary spastic paraplegia mutations affect katanin's conformational transition using molecular simulations, revealing that some mutations impact an intermediate state.

This study uses molecular dynamics and machine learning to analyze how the R591S mutation in spastin affects its dynamics and allostery, leading to impaired microtubule severing and HSP-related dysfunction.

This paper demonstrates a proof-of-concept model using machine learning and SERS to predict cancer from small extracellular vesicles, with explainable AI for interpretability.

This paper introduces StELa, a clustering method that identifies protein conformational minima more effectively than existing methods like GROMOS and CATS.