David T. Jones

This paper evaluates DrugCLIP's contrastive learning embeddings for virtual screening, showing strong performance but limitations under realistic conditions.

TEDlm is a domain-centric protein language model that improves remote-homology detection and Molecular Function prediction by pretraining on structurally-defined domain sequences.

This paper reports on the fourth CAFA challenge, showing improvements in Gene Ontology term prediction but highlighting challenges in new evaluation settings.

The paper shows that febrile heat stress increases cytoadhesion in Plasmodium falciparum-infected red blood cells by enhancing surface protein trafficking, including PfEMP1 and PSAC.

This study investigates the evolution and inhibition of FIKK kinases in P. falciparum, showing they can be inhibited by a single compound.

The paper combines AlphaFold and ESMfold predictions to create a large structural dataset, identifies novel protein folds and domain combinations, and highlights the value of metagenomic data for structural diversity.

The paper introduces TED, a resource that segments and classifies over 370 million domains from the AlphaFold Protein Structure Database, expanding known structural domains and uncovering new structural interactions.

The paper evaluates and optimizes sequence-based gene regulatory deep learning models using a DREAM Challenge dataset, showing improvements through architectural and training strategy refinements.

Merizo is a deep learning method for domain segmentation in protein structures, trained on CATH domains and applied to human proteome models to identify 40,818 putative domains.

This paper introduces a deep generative model for de novo protein design that efficiently samples novel sequences with ordered structures, outperforming existing methods trained on natural sequences.

This paper demonstrates using automatic differentiation to train a coarse-grained force field for proteins, achieving stable structure and dynamics.

The paper uses AlphaFold to design novel proteins with fixed backbones, showing agreement between predicted structures and desired backbones.

The paper evaluates 43 methods for predicting intrinsically disordered regions and binding regions in proteins, finding that deep learning methods outperform older physicochemical approaches but binding prediction remains challenging.

The paper introduces an ultrafast protein structure prediction method that directly outputs backbone coordinates from MSAs, enabling rapid modeling of uncharacterized proteins.

This work introduces FFPred-GAN, a GAN-based method to generate synthetic protein feature samples, improving function prediction accuracy across all three Gene Ontology domains.

The paper introduces a semi-supervised framework for single-sequence protein modelling, achieving high accuracy in secondary structure prediction.

The paper reports results from the CAFA3 challenge, showing improved computational protein function prediction and new annotations for over 1000 genes through experimental screens in C. albicans, P. aureginosa, and Drosophila.

STRING2GO uses deep maxout neural networks to improve protein function prediction from interaction networks, outperforming existing methods in accuracy.

The paper describes the development and application of DeepMetaPSICOV for contact prediction in CASP13, highlighting improvements in alignment generation and performance.

The paper uses word2vec to create semantically meaningful embeddings of protein domains and applies them to identify functional roles of unknown domains.

This paper introduces a multi-task deep neural network for protein function prediction, showing improved accuracy over baseline methods and single-task predictors.