Biology subjects

Durham, J.

Publications and source records attributed to Durham, J..

7 recordsLinked to original sources

bioRxiv2026

Cataloging cysteines in ECOD domains using a protein language model

Cysteine is among the most chemically versatile residues in the proteome, existing in three competing functional states: metal coordination, covalent disulfide bonding, and bioactive free thiols. Although these states can be readily assigned from experimentally determined protein structures using simple geometric criteria, accurately annotating them from predicted structures remains challenging. To bridge the gap between predicted structures and functional interpretation, we developed TriCyP (Tri-state Cysteine Predictor), an efficient two-layer neural network built on ESM-2 protein language model embeddings. On an independent benchmark set, TriCyP achieves near-perfect accuracy (AUROC = 0.99) and outperforms existing approaches for predicting both disulfide bonding and metal coordination. We applied TriCyP to classify 2.7 million cysteine residues across 0.9 million ECOD F70 representative domains. The resulting proteome-scale landscape recapitulates established biological patterns. Cysteines are enriched in eukaryotes: disulfide-bonded states are concentrated in extracellular proteins, and metal-coordinating cysteines peak in nuclear proteins owing to the abundance of zinc-finger transcription factors. We further demonstrate the utility of cysteine-state annotation through two pilot studies. First, predicted disulfide-forming cysteines lacking a corresponding structural partner in AlphaFold models may identify either regions of elevated structural uncertainty or latent inter-protein disulfide bonds that stabilize protein-protein interactions. Second, systematic analysis of known and predicted metal-coordinating cysteines across ECOD homologous groups uncovers previously unrecognized metal-binding protein families. This proteome-wide catalog of cysteine states is available as a community resource (http://prodata.swmed.edu/tricyp) and will be integrated into future ECOD releases.

bioinformatics↗

bioRxiv2025

Benchmarking of Human ReadRemoval Strategies for Viral andMicrobial Metagenomics

Human reads are a key contaminant in microbial metagenomics and enrichment-based studies, requiring removal for computational efficiency, biological analysis, and privacy protection. Various in silico methods exist, but their effectiveness depends on the parameters and reference genomes used. Here, we assess different methods, including the impact of the updated T2T-CHM13 human genome versus GRCh38. Using a synthetic dataset of viral and human reads, we evaluated performance metrics for multiple approaches. We found that the usage of high-sensitivity configuration of Bowtie2 with the T2T-CHM13 reference assembly significantly improves human read removal with minimal loss of specificity, albeit at higher computational cost compared to other methods investigated. Applying this approach to a publicly available microbiome dataset, we effectively removed sex-determining SNPs with little impact on microbial assembly. Our results suggest that our high-sensitivity Bowtie2 approach with the T2T-CHM13 is the best method tested to minimise identifiability risks from residual human reads.

genomics↗

bioRxiv2024

Computing the Human Interactome

Protein-protein interactions (PPI) are essential for biological function. Recent advances in coevolutionary analysis and Deep Learning (DL) based protein structure prediction have enabled comprehensive PPI identification in bacterial and yeast proteomes, but these approaches have limited success to date for the more complex human proteome. Here, we overcome this challenge by 1) enhancing the coevolutionary signals with 7-fold deeper multiple sequence alignments harvested from 30 petabytes of unassembled genomic data, and 2) developing a new DL network trained on augmented datasets of domain-domain interactions from 200 million predicted protein structures. These advancements allow us to systematically screen through 200 million human protein pairs and predict 18,316 PPIs with an expected precision of 90%, among which 5,578 are novel predictions. 3D models of these predicted PPIs nearly triple the number of human PPIs with accurate structural information, providing numerous insights into protein function and mechanisms of human diseases.

biochemistry↗

bioRxiv2023

Ca2+-sensing receptor regulates neuronal excitability via Kv7 channel and Gi/o protein signalling

Neuropathic pain, a debilitating condition with unmet medical needs, can be charactarised as hyperexcitability of nociceptive neurons caused by dysfunction of ion channels. Voltage-gated potassium channel type 7 (Kv7), responsible for maintaining neuronal resting membrane potential and thus neuronal exitability, resides under tight control of G protein-coupled receptors (GPCR). Calcium-sensing receptor (CaSR) is a GPCR that is known to regulate activity of numerous ion channels, but whether CaSR could control Kv7 channel function has been unexplored until now. Our results demonstrate that CaSR is expressed in recombinant cell models, human induced pluripotent stem cell (hiPSC)-derived nociceptive-like neurons and mouse dorsal root ganglia neurons, and its activation induced depolarisation via Kv7.2/7.3 channel inhibition. The CaSR-Kv7.2/7.3 channel crosslink was mediated via the Gi/o protein/adenylate cyclase/cyclic adenosine monophosphate/protein kinase A signalling cascade. Suppression of CaSR function rescued hiPSC-derived nociceptive-like neurons from algogenic cocktail-induced hyperexcitability. To conclude, this study demonstrates that CaSR-Kv7.2/7.3 channel crosslink via the Gi/o protein signalling pathway effectively regulates neuronal excitability, providing a feasible pharmacological target for neuronal hyperexcitability management in neuropathic pain.

physiology↗

bioRxiv2022

DPAM: A Domain Parser for AlphaFold Models

The recent breakthroughs in structure prediction, where methods such as AlphaFold demonstrated near atomic accuracy, herald a paradigm shift in structure biology. The 200 million high-accuracy models released in the AlphaFold Database are expected to guide protein science in the coming decades. Partitioning these AlphaFold models into domains and subsequently assigning them to our evolutionary hierarchy provides an efficient way to gain functional insights of proteins. However, classifying such a large number of predicted structures challenges the infrastructure of current structure classifications, including our Evolutionary Classification of protein Domains (ECOD). Better computational tools are urgently needed to automatically parse and classify domains from AlphaFold models. Here we present a Domain Parser for AlphaFold Models (DPAM) that can automatically recognize globular domains from these models based on predicted aligned errors, inter-residue distances in 3D structures, and ECOD domains found by sequence (HHsuite) and structural (DALI) similarity searches. Based on a benchmark of 18,759 AlphaFold models, we demonstrated that DPAM could recognize 99.5% domains and assign correct boundaries for 85.2% of them, significantly outperforming structure-based domain parsers and homology-based domain assignment using ECOD domains found by HHsuite or DALI. Application of DPAM to the massive set of AlphaFold models will allow for more efficient classification of domains, providing evolutionary contexts and facilitating functional studies.

bioinformatics↗

bioRxiv2022

Towards a structurally resolved cancer interactome

Protein-protein interactions (PPIs) are involved in almost all essential cellular processes. Perturbation of PPI networks plays critical roles in tumorigenesis, cancer progression and metastasis. While numerous high-throughput experiments have produced a vast amount of data for PPIs, these datasets suffer from high false positive rates and exhibit a high degree of discrepancy. Coevolution of amino acid positions between protein pairs has proven to be useful in identifying interacting proteins and providing structural details of the interaction interfaces with the help of deep learning methods like AlphaFold (AF). In this study, we applied AF to investigate the cancer protein-protein interactome. We predicted 1,798 PPIs for cancer driver proteins involved in diverse cellular processes such as transcription regulation, signal transduction, DNA repair and cell cycle. We modeled the spatial structure for the predicted binary protein complexes, 1,087 of which lacked previous 3D structure information. Our predictions offer novel structural insight into many cancer-related processes such as the MAP kinase cascade and Fanconi anemia pathway. We further investigated the cancer mutation landscape by mapping somatic missense mutations (SMMs) in cancer to the predicted PPI interfaces and performing enrichment and depletion analyses. Interfaces enriched or depleted with SMMs exhibit different preferences for functional categories. Interfaces enriched in mutations tend to function in pathways that are deregulated in cancers and they may help explain the molecular mechanisms of cancers in patients; interfaces lacking mutations appear to be essential for the survival of cancer cells and thus may be future targets for PPI modulating drugs.

bioinformatics↗

bioRxiv2020

Evaluating aerosol and splatter following dental procedures

BackgroundDental procedures often produce aerosol and splatter which are potentially high risk for spreading pathogens such as SARS-CoV-2. The existing literature is limited. Objective(s)To develop a robust, reliable and valid methodology to evaluate distribution and persistence of dental aerosol and splatter, including the evaluation of clinical procedures. MethodsFluorescein was introduced into the irrigation reservoirs of a high-speed air-turbine, ultrasonic scaler and 3-in-1 spray and procedures performed on a mannequin in triplicate. Filter papers were placed in the immediate environment. The impact of dental suction and assistant presence were also evaluated. Samples were analysed using photographic image analysis, and spectrofluorometric analysis. Descriptive statistics were calculated and Pearsons correlation for comparison of analytic methods. ResultsAll procedures were aerosol and splatter generating. Contamination was highest closest to the source, remaining high to 1-1.5 m. Contamination was detectable at the maximum distance measured (4 m) for high-speed air-turbine with maximum relative fluorescence units (RFU) being: 46,091 at 0.5 m, 3,541 at 1.0 m, and 1,695 at 4 m. There was uneven spatial distribution with highest levels of contamination opposite the operator. Very low levels of contamination ([≤]0.1% of original) were detected at 30 and 60 minutes post procedure. Suction reduced contamination by 67-75% at 0.5-1.5 m. Mannequin and operator were heavily contaminated. The two analytic methods showed good correlation (r=0.930, n=244, p<0.001). ConclusionDental procedures have potential to deposit aerosol and splatter at some distance from the source, being effectively cleared by 30 minutes in our setting.

microbiology↗