Luc G. Elliott

DPI-score: A deep learning-based metric for assessing protein-protein interfaces in cryo-EM derived assemblies

The paper introduces DPI-Score, a deep learning metric to assess protein-protein interfaces in cryo-EM structures, showing it can detect interface errors not found by density-based scores.

Phosphorylation at S345 Converts HIF-2α from a Transcription Factor to an RNA binding protein

This study shows that phosphorylation at S345 converts HIF-2α from a transcription factor to an RNA-binding protein, altering its function in hypoxia response.

Capturing α-synuclein polymorphism with AlphaFold 3 and rivals

This paper evaluates AlphaFold 3 and other methods in predicting α-synuclein polymorphs using a reference set of known structures and finds they can generate fibril-like models, though performance varies with template use.

AlphaFold 2, but not AlphaFold 3, predicts confident but unrealistic beta-solenoid structures for repeat proteins

The paper evaluates AlphaFold 2's performance on repeat proteins, finding it predicts confident but unrealistic beta-solenoid structures, unlike other methods.

ABCFold: easier running and comparison of AlphaFold 3, Boltz-1 and Chai-1

ABCFold simplifies running and comparing AlphaFold 3, Boltz-1, and Chai-1 for protein structure prediction with standardized input and unified result handling.

Deep-learning protein structure predictions suggest likely molecular functions for three uncharacterised polytopic membrane proteins from the P. falciparum apicoplast

The paper uses AlphaFold2 to predict structures of three uncharacterized apicoplast proteins and suggests their potential molecular functions as transporters.

CASP15 cryoEM protein and RNA targets: refinement and analysis using experimental maps

The paper evaluates CASP15 cryoEM predictions using density maps, showing high accuracy but challenges in side-chain positioning and loop modeling.

Slice'N'Dice: Maximising the value of predicted models for structural biologists

SliceNDice improves Molecular Replacement by splitting and repositioning predicted structural models to enhance accuracy in domain orientations.