Genome-wide prediction of disease variant effects with a deep protein language model

Nature Genetics, Год журнала: 2023, Номер 55(9), С. 1512 - 1522

Опубликована: Авг. 10, 2023

Predicting the effects of coding variants is a major challenge. While recent deep-learning models have improved variant effect prediction accuracy, they cannot analyze all due to dependency on close homologs or software limitations. Here we developed workflow using ESM1b, 650-million-parameter protein language model, predict ~450 million possible missense in human genome, and made predictions available web portal. ESM1b outperformed existing methods classifying ~150,000 ClinVar/HGMD as pathogenic benign predicting measurements across 28 deep mutational scan datasets. We further annotated ~2 damaging only specific isoforms, demonstrating importance considering isoforms when effects. Our approach also generalizes more complex such in-frame indels stop-gains. Together, these results establish an effective, accurate general

Язык: Английский

---

Molecular basis of proton sensing by G protein-coupled receptors

Cell, Год журнала: 2025, Номер unknown

Опубликована: Янв. 1, 2025

Язык: Английский

---

Massively parallel functional testing of MSH2 missense variants conferring Lynch syndrome risk

The American Journal of Human Genetics, Год журнала: 2020, Номер 108(1), С. 163 - 175

Опубликована: Дек. 23, 2020

The lack of functional evidence for the majority missense variants limits their clinical interpretability and poses a key barrier to broad utility carrier screening. In Lynch syndrome (LS), one most highly prevalent cancer syndromes, nearly 90% clinically observed are deemed "variants uncertain significance" (VUS). To systematically resolve status, we performed massively parallel screen in human cells identify loss-of-function DNA mismatch repair factor MSH2. resulting effect map is substantially complete, covering 94% 17,746 possible variants, concordant (96%) with existing data expert clinicians' interpretations. large (89%) were functionally neutral, perhaps unexpectedly light its evolutionary conservation. These provide ready-to-use ∼1,300 extant VUSs MSH2 may facilitate prospective classification newly discovered clinic.

Язык: Английский

---

The full spectrum of SLC22 OCT1 mutations illuminates the bridge between drug transporter biophysics and pharmacogenomics

Molecular Cell, Год журнала: 2024, Номер 84(10), С. 1932 - 1947.e10

Опубликована: Май 1, 2024

Язык: Английский

---

Discovery of non-retinoid compounds that suppress the pathogenic effects of misfolded rhodopsin in a mouse model of retinitis pigmentosa

PLoS Biology, Год журнала: 2025, Номер 23(1), С. e3002932 - e3002932

Опубликована: Янв. 14, 2025

Pathogenic mutations that cause rhodopsin misfolding lead to a spectrum of currently untreatable blinding diseases collectively termed retinitis pigmentosa. Small molecules correct are therefore urgently needed. In this study, we utilized virtual screening search for drug-like bind the orthosteric site rod opsin and improve its folding trafficking. We identified validated biological effects 2 non-retinoid compounds with favorable pharmacological properties cross blood-retina barrier. These reversibly unliganded opsin, each Kd comparable 9-cis-retinal stability. By improving internal protein structure network (PSN), these ligands also enhanced plasma membrane expression total 36 123 tested clinical RP variants, including most prevalent P23H variant. Importantly, protected retinas against light-induced degeneration in mice vulnerable bright light injury prolonged survival photoreceptors pigmentosa mouse model misfolding.

Язык: Английский

---

General trends in the effects of VX-661 and VX-445 on the plasma membrane expression of clinical CFTR variants

Cell chemical biology, Год журнала: 2023, Номер 30(6), С. 632 - 642.e5

Опубликована: Май 29, 2023

Cystic fibrosis (CF) is caused by mutations that compromise the expression and/or function of cystic transmembrane conductance regulator (CFTR) chloride channel. Most people with CF harbor a common misfolded variant (ΔF508) can be partially rescued therapeutic "correctors" restore its expression. Nevertheless, many other variants are insensitive to correctors. Using deep mutational scanning, we quantitatively compare effects two correctors on plasma membrane 129 variants. Though structural calculations suggest corrector binding provides similar stabilization most variants, it's those intermediate and near pockets exhibit greatest response. Deviations in sensitivity appear depend degree destabilization timing misassembly. Combining appears rescue more doubling energy stabilizing distinct cotranslational folding transitions. These results provide an overview rare establish new tools for precision pharmacology.

Язык: Английский

---

Molecular basis of proton-sensing by G protein-coupled receptors

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Апрель 18, 2024

Abstract Three proton-sensing G protein-coupled receptors (GPCRs), GPR4, GPR65, and GPR68, respond to changes in extracellular pH regulate diverse physiology are implicated a wide range of diseases. A central challenge determining how protons activate these is identifying the set residues that bind protons. Here, we determine structures each receptor understand spatial arrangement putative proton sensing active state. With newly developed deep mutational scanning approach, determined functional importance every residue activation for GPR68 by generating ∼9,500 mutants measuring effects on signaling surface expression. This unbiased screen revealed that, unlike other proton-sensitive cell channels receptors, no single site critical recognition GPR68. Instead, network titratable extend from transmembrane region converge canonical class GPCR motifs GPCRs. More broadly, our approach integrating structure interrogation defines new framework understanding rich complexity signaling. One-sentence summary The protonation networks governing human pH-sensing GPCRs uncovered integrative cryo-EM scanning.

Язык: Английский

---

Genome-wide prediction of disease variants with a deep protein language model

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2022, Номер unknown

Опубликована: Авг. 26, 2022

Abstract Distinguishing between damaging and neutral missense variants is an ongoing challenge in human genetics, with profound implications for clinical diagnosis, genetic studies protein engineering. Recently, deep-learning models have achieved state-of-the-art performance classifying as pathogenic or benign. However, these are currently unable to provide predictions over all variants, either because of dependency on close homologs due software limitations. Here we leveraged ESM1b, a 650M-parameter language model, predict the functional impact coding variation at scale. To overcome existing technical limitations, developed modified ESM1b workflow functionalized, first time, proteins genome, resulting ∼450M possible variant effects. was able distinguish benign across ∼150K annotated ClinVar HGMD, outperforming methods. also exceeded state art predicting experimental results deep mutational scans. We further ∼2M ∼9K alternatively-spliced genes certain isoforms while others, demonstrating importance considering when functionalizing The complete catalog effect available at: https://huggingface.co/spaces/ntranoslab/esm_variants .

Язык: Английский

---

Evolutionary history of ATP‐binding cassette proteins

FEBS Letters, Год журнала: 2020, Номер 594(23), С. 3882 - 3897

Опубликована: Ноя. 4, 2020

ATP-binding cassette (ABC) proteins are found in every sequenced genome and evolved deep the phylogenetic tree of life. ABC form one largest homologous protein families, with most being involved substrate transport across biological membranes, a few cytoplasmic members regulating essential processes like translation. The predominant classification scheme is derived from human members, but increasing number fully genomes permits to reevaluate this paradigm light evolutionary history ABC-protein superfamily. As we study diversity substrates, mechanisms, physiological roles proteins, knowledge relationships highlights similarities differences that can be attributed specific branches divergence. While alignments trees built on natural sequence variation account for divergence high-throughput experiments next-generation sequencing creating experimental instrumental identifying functional constraints. combination experimentally produced allows broader more rational function proteins.

Язык: Английский

---

Computational modeling and prediction of deletion mutants

Structure, Год журнала: 2023, Номер 31(6), С. 713 - 723.e3

Опубликована: Апрель 28, 2023

Язык: Английский

---