Fig. 1
From: End-to-end protein–ligand complex structure generation with diffusion-based generative models
Overview of the proposed framework. For inputs, protein amino acid sequence and ligand molecular graph are employed. The conformational sampling process involves the iterative application of input featurization, residual feature update, and equivariant denoising to generate an ensemble of complex structures. ESM-2 [18], a large-scale protein language model (PLM), is utilized to featurize input protein sequences