MolPIF: A Parameter Interpolation Flow Model for Molecule Generation
This work addresses the need for more flexible and efficient generative models in drug discovery, offering incremental improvements over existing parameter-space-based methods.
The paper tackled the limitations of Bayesian Flow Networks in molecular generation by proposing a Parameter Interpolation Flow model, which demonstrated superior performance across diverse metrics for structure-based drug design.
Advances in deep learning for molecular generation show promise in accelerating drug discovery. Bayesian Flow Networks (BFNs) have recently shown impressive performance across diverse chemical tasks, with their success often ascribed to the paradigm of modeling in a low-variance parameter space. However, the Bayesian inference-based strategy imposes limitations on designing more flexible distribution transformation pathways, making it challenging to adapt to diverse data distributions and varied task requirements. Furthermore, the potential for simpler, more efficient parameter-space-based models is unexplored. To address this, we propose a novel Parameter Interpolation Flow model (named PIF) with detailed theoretical foundation, training, and inference procedures. We then develop MolPIF for structure-based drug design, demonstrating its superior performance across diverse metrics compared to baselines. This work validates the effectiveness of parameter-space-based generative modeling paradigm for molecules and offers new perspectives for model design.