John E. Pak

Language Modeling Materializes a World Model of Protein Biology

The paper uses language modeling to create a comprehensive representation of protein biology, enabling structure prediction and discovery of high-affinity protein interactions.

Therapeutic Potential of Cyclodextrins Targeting Dengue Virus and SARS-CoV-2 Infection and Pathogenesis

This study evaluates cyclodextrins' ability to reduce endothelial dysfunction and antiviral activity against DENV and SARS-CoV-2 in vitro and in vivo.

ADAR1 Restricts Poliovirus Infection through dual disruption of translation initiation and protein coding

The paper identifies ADAR1 as a host factor that restricts poliovirus infection by editing viral RNA to disrupt translation initiation and introduce a coding error.

The Interplay of Furin Cleavage and D614G in Modulating SARS-CoV-2 Spike Protein Dynamics

The paper analyzes SARS-CoV-2 spike protein dynamics using eVLPs and HDX-MS, showing the D614G mutation and furin cleavage influence conformational states and infectivity.

The Virtual Lab: AI Agents Design New SARS-CoV-2 Nanobodies with Experimental Validation

Role-specialised LLM agents under an LLM principal investigator, with light human steering, assembled an ESM/AlphaFold-Multimer/Rosetta pipeline and designed 92 SARS-CoV-2 nanobodies; two bound recent variants better.

High-yield production of recombinant human myelin oligodendrocyte glycoprotein in SHuffle bacteria without a refolding step

This paper describes a method to produce soluble recombinant human MOG in SHuffle bacteria without refolding, enabling B cell-dependent EAE studies.

CCDC32 stabilizes clathrin-coated pits and drives their invagination

This study identifies CCDC32 as an endocytic accessory protein that stabilizes clathrin-coated pits and drives their invagination, with mutations linked to cardio-facio-neuro-developmental syndrome.

Efficient evolution of human antibodies from general protein language models and sequence information alone

Consensus likelihoods from general, unsupervised protein language models evolve seven human antibodies using only a wildtype sequence and <=20 tested variants, improving binding affinity up to 160-fold.