Junsu Ko

OmniScore: Universal Scoring of Diverse Biomolecular Complexes via Equivariant Geometry-Aware Discrete Representation Learning

OmniScore learns shared geometric representations of diverse biomolecular complexes and improves quality assessment and interaction prediction across molecular classes.

Benchmarking Sequence-Based and AlphaFold-Based Methods for pMHC-II Binding Core Prediction: Distinct Strengths and Consensus Approaches

This paper benchmarks sequence-based and AlphaFold-based methods for predicting pMHC-II binding cores, finding that AlphaFold-based methods excel at identifying binders while NetMHCIIpan is better at identifying non-binders.

STRUMP-I: Structure-based machine learning approach to pMHC-I binding prediction using force field energy features

STRUMP-I improves pMHC-I binding prediction by using force-field energy features, outperforming sequence-based methods on underrepresented MHC alleles and showing synergy with them.

De novo design of a safe and potent respiratory syncytial virus immuno-focusing antigen

This study designs a de novo RSV antigen using AI to reduce immunopathology while maintaining immunogenicity, showing safety and efficacy in preclinical models.

Allosteric deimmunisation of a Salmonella phosphatase enhances catalytic function

This study deimmunizes a Salmonella phosphatase by mutating buried residues, enhancing its catalytic function while reducing immunogenicity.

Applying multi-state modeling using AlphaFold2 for kinases and its application for ensemble screening

This paper introduces a multi-state modeling protocol for kinases using AlphaFold2 to improve structure-based virtual screening by overcoming structural biases and enhancing the identification of diverse inhibitors.

A-Prot: Protein structure modeling using MSA transformer

A-Prot uses MSA Transformer to predict protein 3D structures, outperforming existing methods in long-range contact prediction and CASP14 target modeling.

Can AlphaFold2 predict protein-peptide complex structures accurately?

Concatenating receptor and peptide sequences via a poly-glycine linker enables AlphaFold2 to predict bound peptide conformations within 3.0 A C-RMSD for over half of benchmarked protein-peptide complexes.