Masahito Ohue

ALLM-Ab uses fine-tuned protein language models and active learning to optimize antibody sequences for both high binding affinity and developability.

The paper improves structure-based virtual screening by modifying AlphaFold2-predicted protein structures to better capture ligand-induced conformations, enhancing screening accuracy.

This study benchmarks HelixFold3's ability to predict holo structures for free energy perturbation calculations, showing improved performance over AlphaFold2 and existing methods.

This paper improves a method for predicting protein-ligand affinity by replacing deep learning with Jensen-Shannon divergence, reducing computation time and improving correlation sign prediction.

This study shows that using AlphaFold3 with ligand inputs improves virtual screening performance compared to apo structures, with better results for low molecular weight ligands.

SpatialPPI 2.0 predicts protein-protein interactions using graph neural networks and inter-residue analysis, outperforming existing methods and working with AlphaFold-predicted structures.

SpatialPPI predicts protein-protein interactions using 3D structural data from AlphaFold Multimer, outperforming existing methods in benchmark tests.

The paper designs cyclic peptides targeting protein-protein interactions using AlphaFold with a cyclic peptide complex offset, achieving high accuracy in structure prediction and binding energy.

This paper evaluates AlphaFold2 and binder hallucination for designing high-affinity antibodies by redesigning sequences of antigen-antibody complexes.

The paper designs solubility-aware peptides targeting protein-protein interactions using a modified AlphaFold-based method, showing improved solubility and binding affinity.

The paper introduces a re-ranking method for protein-peptide docking using amino acid contact profiles from rigid-body docking results, showing improved prediction accuracy.

The paper improves protein-protein docking by combining consensus-based rescoring with the ZDOCK native scoring function, showing that interface size information is critical for better results.