Combining evolution and protein language models for an interpretable cancer driver mutation prediction with D2Deep DOI Creative Commons
Konstantina Tzavella, Adrián Díaz, Catharina Olsen

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 17, 2023

Abstract The mutations driving cancer are being increasingly exposed through tumor-specific genomic data. However, differentiating between cancer-causing driver and random passenger remains challenging. State-of-the-art homology-based predictors contain built-in biases often ill-suited to the intricacies of biology. Protein Language Models have successfully addressed various biological problems but not yet been tested on challenging task mutation prediction at large scale. Additionally, they fail offer result interpretation, hindering their effective use in clinical settings. AI-based D2Deep method we introduce here addresses these challenges by combining two powerful elements: i) a non-specialized protein language model that captures makeup all sequences ii) protein-specific evolutionary information encompasses functional requirements for particular protein. relies exclusively sequence information, outperforms state-of-the-art intricate epistatic changes throughout caused mutations. These correlate with known setting can be used interpretation results. is trained balanced, somatic training set so effectively mitigates related hotspot compared techniques. versatility illustrated its performance non-cancer prediction, where most variants still lack consequences. predictions confidence scores available via https://tumorscope.be/d2deep help prioritization.

Language: Английский

Decoding Phospho-Regulation and Flanking Regions in Autophagy-Associated Short Linear Motifs: A Case Study of Optineurin-LC3B Interaction DOI Creative Commons
Oana N. Antonescu, Mattia Utichi, Valentina Sora

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Sept. 30, 2023

Abstract Short Linear Motifs (SLiMs) play a pivotal role in mediating interactions between intrinsically disordered proteins and their binding partners. SLiMs exhibit sequence degeneracy undergo regulation through post-translational modifications, including phosphorylation. The flanking regions surrounding the core motifs also exert crucial shaping modes of interaction. In this study, we aimed to integrate biomolecular simulations, silico high-throughput mutational scans, biophysical experiments elucidate structural details phospho-regulation class for autophagy, known as LC3 interacting (LIRs). As case investigated interaction optineurin LC3B. Optineurin LIR perfectly exemplify where there is complex interplay different phosphorylations N-terminal helical region be disentangled. Our work unveils unexplored upstream motif contributing interface. results offer an atom-level perspective on mechanisms conformational alterations induced by phosphorylation LC3B recognition, along with effects mutations background phosphorylated form protein. Additionally, assessed impact disease-related optineurin, accounting functional features. Notably, established approach based Microfluidic Diffusional Sizing novel method investigate affinity target proteins, enabling precise measurements dissociation constant selection variants identified screening. Overall, our provides versatile toolkit characterize other LIR-containing modulation or phospho-regulated SLiMs, thereby advancing understanding important cellular processes.

Language: Английский

Citations

1
AVENGERS: Analysis of Variant Effects using Next Generation sequencing to EnhanceBRCA2Stratification DOI Open Access
Sounak Sahu,

Melissa Galloux,

Eileen Southon

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Dec. 15, 2023

Abstract Accurate interpretation of genetic variation is a critical step towards realizing the potential precision medicine. Sequencing-based tests have uncovered vast array BRCA2 sequence variants. Due to limited clinical, familial and/or epidemiological data, thousands variants are considered be uncertain significance (VUS). To determine functional impact VUSs, here we develop AVENGERS: Analysis Variant Effects using NGs Enhance Stratification, utilizing CRISPR-Cas9-based saturation genome editing (SGE) in humanized-mouse embryonic stem cell line. We categorized nearly all possible missense single nucleotide (SNVs) encompassing C-terminal DNA binding domain BRCA2. generated function scores for 6270 SNVs, covering 95.5% SNVs exons 15-26 spanning residues 2479-3216, including 1069 unique VUS, with 81% and 14% found nonfunctional. Our classification aligns strongly pathogenicity data from ClinVar, orthogonal assays computational meta predictors. statistical classifier exhibits 92.2% sensitivity 96% specificity distinguishing clinically benign pathogenic recorded ClinVar. Furthermore, offer proactive evidence 617 being non-functional 3396 demonstrated by on growth response damaging drugs like cisplatin olaparib. This serves as valuable resource interpreting unidentified population physicians counselors assessing VUSs patients.

Language: Английский

Citations

1
TRAP1S-nitrosylation as a model of population-shift mechanism to study the effects of nitric oxide on redox-sensitive oncoproteins DOI Creative Commons
Elena Papaleo, Matteo Tiberti,

Matteo Arnaudi

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: Dec. 11, 2022

Abstract S -nitrosylation is a post-translational modification in which nitric oxide (NO) binds to the thiol group of cysteine, generating an -nitrosothiol (SNO) adduct. has different physiological roles, and its alteration also been linked growing list pathologies, including cancer. SNO can affect function stability proteins, such as mitochondrial chaperone TRAP1. Interestingly, site (C501) TRAP1 proximity another cysteine (C527). This feature suggests that -nitrosylated C501 could engage disulfide bridge with C527 TRAP1, resembling well-known ability cysteines resolve vicinal cysteines. We used enhanced sampling simulations in-vitro biochemical assays address structural mechanisms induced by S- nitrosylation. showed induces conformational changes proximal favors conformations suitable for disulfide-bridge formation. explored 4172 known proteins using high-throughput analyses. Furthermore, we carried out coarse-grain 44 account protein dynamics resulted identification up 1248 examples sense redox state site, opening new perspectives on biological effects switches. In addition, devised two bioinformatic workflows ( https://github.com/ELELAB/SNO_investigation_pipelines ) identify or accompanying annotations. Finally, analyzed mutations tumor suppressor oncogenes connection switch -nitrosylation. classified variants neutral, stabilizing, destabilizing respect propensity be undergo population-shift mechanism. The methods applied here provide comprehensive toolkit future studies candidates, variant classification, rich data source research community NO field.

Language: Английский

Citations

1
Combining evolution and protein language models for an interpretable cancer driver mutation prediction with D2Deep DOI Creative Commons
Konstantina Tzavella, Adrián Díaz, Catharina Olsen

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 17, 2023

Abstract The mutations driving cancer are being increasingly exposed through tumor-specific genomic data. However, differentiating between cancer-causing driver and random passenger remains challenging. State-of-the-art homology-based predictors contain built-in biases often ill-suited to the intricacies of biology. Protein Language Models have successfully addressed various biological problems but not yet been tested on challenging task mutation prediction at large scale. Additionally, they fail offer result interpretation, hindering their effective use in clinical settings. AI-based D2Deep method we introduce here addresses these challenges by combining two powerful elements: i) a non-specialized protein language model that captures makeup all sequences ii) protein-specific evolutionary information encompasses functional requirements for particular protein. relies exclusively sequence information, outperforms state-of-the-art intricate epistatic changes throughout caused mutations. These correlate with known setting can be used interpretation results. is trained balanced, somatic training set so effectively mitigates related hotspot compared techniques. versatility illustrated its performance non-cancer prediction, where most variants still lack consequences. predictions confidence scores available via https://tumorscope.be/d2deep help prioritization.

Language: Английский

Citations

0