Prediction of Recurrent Mutations in SARS-CoV-2 Using Artificial Neural Networks

Saldivar-Espinoza, Bryan; Macip, Guillem; Garcia-Segura, Pol; Mestres-Truyol, Júlia; Puigbò, Pere; Ceretó-Massagué, Adrià; Pujadas, Gerard; Garcı́a-Vallvé, Santiago

doi:10.3390/ijms232314683

Cited by 11 publications

(12 citation statements)

References 78 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The PHATE visualization of the 296,437 retained de novo iSNVs by genomic position display no clustering according to the mutational pattern, underscoring the optimal curation of the dataset (Figure 6A). Additionally, the substitution spectrum of this curated set shows a prevalence of C > T and G > T substitutions (Figure 6B), aligning with consensus iSNV patterns and known SARS-CoV-2 mutational trends (Moshiri et al 2023; Fumagalli et al 2023; Bloom et al 2023; Saldivar-Espinoza et al 2023).…”

Section: Resultssupporting

confidence: 71%

Refining SARS-CoV-2 Intra-host Variation by Leveraging Large-scale Sequencing Data

Mostefai,

Grenier,

Poujol

et al. 2024

Preprint

View full text Add to dashboard Cite

Understanding the evolution of viral genomes is essential for elucidating how viruses adapt and change over time. Analyzing intra-host single nucleotide variants (iSNVs) provides key insights into the mechanisms driving the emergence of new viral lineages, which is crucial for predicting and mitigating future viral threats. Despite the potential of next-generation sequencing (NGS) to capture these iSNVs, the process is fraught with challenges, particularly the risk of capturing sequencing artifacts that may result in false iSNVs. To tackle this issue, we developed a two-step workflow designed to enhance the reliability of iSNV detection in large heterogeneous collections of NGS libraries. We use over 130,000 publicly available SARS-CoV-2 NGS libraries to show how our comprehensive workflow effectively distinguishes emerging viral mutations from sequencing errors. This approach incorporates rigorous bioinformatics protocols, stringent quality control metrics, and innovative usage of dimensionality reduction methods to generate insightful representations of this high-dimensional dataset. We identified and mitigated notable batch effects linked to specific sequencing centers around the world and introduced quality control metrics such as the Strand Bias Likelihood that considers strand coverage imbalance, enhancing iSNV reliability. Additionally, we pioneer the application of the PHATE visualization approach to genomic data and introduce a methodology that quantifies how closely related groups of data points are within a two-dimensional space, enhancing our ability to explain clustering patterns based on their shared genetic characteristics. Our workflow not only sheds light on the complexities of viral genomic analysis with state-of-the-art sequencing technologies but also advances the detection of accurate intra-host mutations, opening the door for an enhanced understanding of viral adaptation mechanisms.

show abstract

Section: Resultssupporting

confidence: 71%

Refining SARS-CoV-2 Intra-host Variation by Leveraging Large-scale Sequencing Data

Mostefai,

Grenier,

Poujol

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…This is mainly due to the fact that SARS-CoV-2 evolution through accumulation of escape genotypes is catalyzed by the presence of human anti-SARS-CoV-2 antibodies elicited by natural or artificial immunity [ 28 ]. Notably, the potential suitability of this proof-of concept study has been recently confirmed by another article, which used artificial neural networks for predicting recurrent mutations (rather than variants) in SARS-CoV-2 [ 29 ]. Briefly, the authors used a machine approach based on genomic rather than ecologic variables for identifying positions in SARS-CoV-2 genome where recurrent mutations are more likely to occur, ultimately achieving 0.79 sensitivity, 0.69 specificity and an overall area under the curve (AUC) of 0.80.…”

Section: Discussionmentioning

confidence: 91%

“…The continued evolution and almost unremittent emergence of new dominant and phylogenetically divergent SARS-CoV-2 Omicron sublineages presents a public health dilemma (i.e., a so-called “permacrisis”) and reinforces the need to develop predictive tools that may be capable of quickly identifying the escape of new variants from previous immunity. The feasibility and reliability of predicting the emergence of recurrent SARS-CoV-2 mutations has been recently demonstrated in the work of Saldivar-Espinoza et al [ 29 ], who developed a very complex mathematical model which was capable of anticipating nucleotide mutations and RNA reactivity with around 80% accuracy. We planned a similar approach for anticipating the emergence and spread of new variants of concerns, which does not detect divergence from a single nucleotide levels, but simplifies the entire process by including simple epidemiologic variables, such as date of emergence of new variants and number of COVID-19 cases and vaccinations, neither of which require complicated calculations or specific software.…”

Section: Discussionmentioning

confidence: 99%

A Simple Epidemiologic Model for Predicting Impaired Neutralization of New SARS-CoV-2 Variants

2023

View full text Add to dashboard Cite

This study is aimed at developing a simple epidemiologic model that could help predict the impaired neutralization of new SARS-CoV-2 variants. We explored the potential association between neutralization of recent and more prevalent SARS-CoV-2 sublineages belonging to the Omicron family (i.e., BA.4/5, BA.4.6, BA.2.75.2, BQ.1.1 and XBB.1) expressed as FFRNT50 (>50% suppression of fluorescent foci fluorescent focus reduction neutralization test) in recipients of four doses of monovalent mRNA-based coronavirus disease 2019 (COVID-19) vaccines, with epidemiologic variables like emergence date and number of spike protein mutations of these sublineages, cumulative worldwide COVID-19 cases and cumulative number of COVID-19 vaccine doses administered worldwide at the time of SARS-CoV-2 Omicron sublineage emergence. In the univariate analysis, the FFRNT50 value for the different SARS-CoV-2 Omicron sublineages was significantly associated with all such variables except with the number of spike protein mutations. Such associations were confirmed in the multivariate analysis, which enabled the construction of the equation: “−0.3917 × [Emergence (date)] + 1.403 × [COVID-19 cases (million)] − 121.8 × [COVID-19 Vaccine doses (billion)] + 18,250”, predicting the FFRNT50 value of the five SARS-CoV-2 Omicron sublineages with 0.996 accuracy (p = 0.013). We have shown in this work that a simple mathematical approach, encompassing a limited number of widely available epidemiologic variables, such as emergence date of new variants and number of COVID-19 cases and vaccinations, could help identifying the emergence and surge of future lineages with major propensity to impair humoral immunity.

show abstract

“…The SARS-CoV-2 pandemic has motivated the collection of virus genomic sequences on an unprecedented scale, which has generated invaluable data on the genomic diversity of an RNA virus. From the ensemble of observed consensus sequences of infected hosts we can extract, for the first time, an exhaustive map of possible amino acid replacements in viral proteins that are tolerable for viable virus (Zhao et al 2022 ;Saldivar-Espinoza et al 2023 ). This brings into stark relief our limited understanding of the genotype/phenotype relationship, which is very detailed on some local functional aspects, such as spike protein antigenicity, but not much developed in general.…”

Section: Discussionmentioning

confidence: 99%

Modulation of biophysical properties of nucleocapsid protein in the mutant spectrum of SARS-CoV-2

Nguyen,

Zhao,

Myagmarsuren

et al. 2024

eLife

View full text Add to dashboard Cite

Genetic diversity is a hallmark of RNA viruses and the basis for their evolutionary success. Taking advantage of the uniquely large genomic database of SARS-CoV-2, we examine the impact of mutations across the spectrum of viable amino acid sequences on the biophysical phenotypes of the highly expressed and multifunctional nucleocapsid protein. We find variation in the physicochemical parameters of its extended intrinsically disordered regions (IDRs) sufficient to allow local plasticity, but also exhibiting functional constraints that similarly occur in related coronaviruses. In biophysical experiments with several N-protein species carrying mutations associated with major variants, we find that point mutations in the IDRs can have nonlocal impact and modulate thermodynamic stability, secondary structure, protein oligomeric state, particle formation, and liquid-liquid phase separation. In the Omicron variant, distant mutations in different IDRs have compensatory effects in shifting a delicate balance of interactions controlling protein assembly properties, and include the creation of a new protein-protein interaction interface in the N-terminal IDR through the defining P13L mutation. A picture emerges where genetic diversity is accompanied by significant variation in biophysical characteristics of functional N-protein species, in particular in the IDRs.

show abstract

Prediction of Recurrent Mutations in SARS-CoV-2 Using Artificial Neural Networks

Cited by 11 publications

References 78 publications

Refining SARS-CoV-2 Intra-host Variation by Leveraging Large-scale Sequencing Data

Refining SARS-CoV-2 Intra-host Variation by Leveraging Large-scale Sequencing Data

A Simple Epidemiologic Model for Predicting Impaired Neutralization of New SARS-CoV-2 Variants

Modulation of biophysical properties of nucleocapsid protein in the mutant spectrum of SARS-CoV-2

Contact Info

Product

Resources

About