Learning from prepandemic data to forecast viral escape

Computational Analysis of the Accumulation of Mutations in Therapeutically Important RNA Viral Proteins During Pandemics with Special Emphasis on SARS-CoV-2

Single stranded RNA viruses are primary causative agents for pandemics, causing extensive morbidity and mortality worldwide. A pivotal question in pandemic preparedness and therapeutic intervention is what are the specific mutations which are more likely to emerge during such global health crises? This study aims to identify markers for mutations with the highest probability of emergence in these pandemics, focusing on the SARS-CoV-2 spike protein, an essential and therapeutically significant viral protein, starting from sequence information from the onset of the pandemic until July 2022. Quite consistently, we observed that emerged mutations tended to demonstrate a high genetic score, which reflects high similarity of the type of codon required for translation between an amino acid and to the mutated one. Further, this pattern is also observed in therapeutically significant proteins of other ssRNA pandemic viruses, including influenza (HA, NA), spike proteins of Ebola, envelope of Dengue and Chikungunya. We propose that the genetic score serves as an initial indicator, preceding the actual impact of the mutation on viral fitness. Finally, we developed a comprehensive computational pipeline to further explore and predict the subsequent effects of mutations on viral fitness. We believe that our pipeline can narrow down and predict future mutations in therapeutically important viral proteins during a pandemic.

Enhancing predictions of protein stability changes induced by single mutations using MSA-based Language Models

MOTIVATION

Protein Language Models offer a new perspective for addressing challenges in structural biology, while relying solely on sequence information. Recent studies have investigated their effectiveness in forecasting shifts in thermodynamic stability caused by single amino acid mutations, a task known for its complexity due to the sparse availability of data, constrained by experimental limitations. To tackle this problem, we introduce two key novelties: leveraging a Protein Language Model that incorporates Multiple Sequence Alignments to capture evolutionary information, and using a recently released mega-scale dataset with rigorous data pre-processing to mitigate overfitting.

RESULTS

We ensure comprehensive comparisons by fine-tuning various pre-trained models, taking advantage of analyses such as ablation studies and baselines evaluation. Our methodology introduces a stringent policy to reduce the widespread issue of data leakage, rigorously removing sequences from the training set when they exhibit significant similarity with the test set. The MSA Transformer emerges as the most accurate among the models under investigation, given its capability to leverage co-evolution signals encoded in aligned homologous sequences. Moreover, the optimized MSA Transformer outperforms existing methods and exhibits enhanced generalization power, leading to a notable improvement in predicting changes in protein stability resulting from point mutations.

AVAILABILITY AND IMPLEMENTATION

Code and data at https://github.com/RitAreaSciencePark/PLM4Muts.

SUPPLEMENTARY INFORMATION

Supplementary Information is available at Bioinformatics online.

HERMES: Holographic Equivariant neuRal network model for Mutational Effect and Stability prediction

Predicting the stability and fitness effects of amino acid mutations in proteins is a cornerstone of biological discovery and engineering. Various experimental techniques have been developed to measure mutational effects, providing us with extensive datasets across a diverse range of proteins. By training on these data, traditional computational modeling and more recent machine learning approaches have advanced significantly in predicting mutational effects. Here, we introduce HERMES, a 3D rotationally equivariant structure-based neural network model for mutational effect and stability prediction. Pre-trained to predict amino acid propensity from its surrounding 3D structure, HERMES can be fine-tuned for mutational effects using our open-source code. We present a suite of HERMES models, pre-trained with different strategies, and fine-tuned to predict the stability effect of mutations. Benchmarking against other models shows that HERMES often outperforms or matches their performance in predicting mutational effect on stability, binding, and fitness. HERMES offers versatile tools for evaluating mutational effects and can be fine-tuned for specific predictive objectives.

MuToN Quantifies Binding Affinity Changes upon Protein Mutations by Geometric Deep Learning

Assessing changes in protein-protein binding affinity due to mutations helps understanding a wide range of crucial biological processes within cells. Despite significant efforts to create accurate computational models, predicting how mutations affect affinity remains challenging due to the complexity of the biological mechanisms involved. In the present work, a geometric deep learning framework called MuToN is introduced for quantifying protein binding affinity change upon residue mutations. The method, designed with geometric attention networks, is mechanism-aware. It captures changes in the protein binding interfaces of mutated complexes and assesses the allosteric effects of amino acids. Experimental results highlight MuToN's superiority compared to existing methods. Additionally, MuToN's flexibility and effectiveness are illustrated by its precise predictions of binding affinity changes between SARS-CoV-2 variants and the ACE2 complex.

HERMES: Holographic Equivariant neuRal network model for Mutational Effect and Stability prediction

Predicting the stability and fitness effects of amino acid mutations in proteins is a cornerstone of biological discovery and engineering. Various experimental techniques have been developed to measure mutational effects, providing us with extensive datasets across a diverse range of proteins. By training on these data, traditional computational modeling and more recent machine learning approaches have advanced significantly in predicting mutational effects. Here, we introduce HERMES, a 3D rotationally equivariant structure-based neural network model for mutational effect and stability prediction. Pre-trained to predict amino acid propensity from its surrounding 3D structure, HERMES can be fine-tuned for mutational effects using our open-source code. We present a suite of HERMES models, pre-trained with different strategies, and fine-tuned to predict the stability effect of mutations. Benchmarking against other models shows that HERMES often outperforms or matches their performance in predicting mutational effect on stability, binding, and fitness. HERMES offers versatile tools for evaluating mutational effects and can be fine-tuned for specific predictive objectives.

Spike deep mutational scanning helps predict success of SARS-CoV-2 clades

SARS-CoV-2 variants acquire mutations in the spike protein that promote immune evasion1 and affect other properties that contribute to viral fitness, such as ACE2 receptor binding and cell entry2,3. Knowledge of how mutations affect these spike phenotypes can provide insight into the current and potential future evolution of the virus. Here we use pseudovirus deep mutational scanning4 to measure how more than 9,000 mutations across the full XBB.1.5 and BA.2 spikes affect ACE2 binding, cell entry or escape from human sera. We find that mutations outside the receptor-binding domain (RBD) have meaningfully affected ACE2 binding during SARS-CoV-2 evolution. We also measure how mutations to the XBB.1.5 spike affect neutralization by serum from individuals who recently had SARS-CoV-2 infections. The strongest serum escape mutations are in the RBD at sites 357, 420, 440, 456 and 473; however, the antigenic effects of these mutations vary across individuals. We also identify strong escape mutations outside the RBD; however, many of them decrease ACE2 binding, suggesting they act by modulating RBD conformation. Notably, the growth rates of human SARS-CoV-2 clades can be explained in substantial part by the measured effects of mutations on spike phenotypes, suggesting our data could enable better prediction of viral evolution.

Machine learning integrative approaches to advance computational immunology

The study of immunology, traditionally reliant on proteomics to evaluate individual immune cells, has been revolutionized by single-cell RNA sequencing. Computational immunologists play a crucial role in analysing these datasets, moving beyond traditional protein marker identification to encompass a more detailed view of cellular phenotypes and their functional roles. Recent technological advancements allow the simultaneous measurements of multiple cellular components-transcriptome, proteome, chromatin, epigenetic modifications and metabolites-within single cells, including in spatial contexts within tissues. This has led to the generation of complex multiscale datasets that can include multimodal measurements from the same cells or a mix of paired and unpaired modalities. Modern machine learning (ML) techniques allow for the integration of multiple "omics" data without the need for extensive independent modelling of each modality. This review focuses on recent advancements in ML integrative approaches applied to immunological studies. We highlight the importance of these methods in creating a unified representation of multiscale data collections, particularly for single-cell and spatial profiling technologies. Finally, we discuss the challenges of these holistic approaches and how they will be instrumental in the development of a common coordinate framework for multiscale studies, thereby accelerating research and enabling discoveries in the computational immunology field.

Mutations in the SARS-CoV-2 spike receptor binding domain and their delicate balance between ACE2 affinity and antibody evasion

Intensive selection pressure constrains the evolutionary trajectory of SARS-CoV-2 genomes and results in various novel variants with distinct mutation profiles. Point mutations, particularly those within the receptor binding domain (RBD) of SARS-CoV-2 spike (S) protein, lead to the functional alteration in both receptor engagement and monoclonal antibody (mAb) recognition. Here, we review the data of the RBD point mutations possessed by major SARS-CoV-2 variants and discuss their individual effects on ACE2 affinity and immune evasion. Many single amino acid substitutions within RBD epitopes crucial for the antibody evasion capacity may conversely weaken ACE2 binding affinity. However, this weakened effect could be largely compensated by specific epistatic mutations, such as N501Y, thus maintaining the overall ACE2 affinity for the spike protein of all major variants. The predominant direction of SARS-CoV-2 evolution lies neither in promoting ACE2 affinity nor evading mAb neutralization but in maintaining a delicate balance between these two dimensions. Together, this review interprets how RBD mutations efficiently resist antibody neutralization and meanwhile how the affinity between ACE2 and spike protein is maintained, emphasizing the significance of comprehensive assessment of spike mutations.

Mutation prediction in the SARS-CoV-2 genome using attention-based neural machine translation

Severe acute respiratory syndrome coronavirus 2 (SARS‑CoV‑2) has been evolving rapidly after causing havoc worldwide in 2020. Since then, it has been very hard to contain the virus owing to its frequently mutating nature. Changes in its genome lead to viral evolution, rendering it more resistant to existing vaccines and drugs. Predicting viral mutations beforehand will help in gearing up against more infectious and virulent versions of the virus in turn decreasing the damage caused by them. In this paper, we have proposed different NMT (neural machine translation) architectures based on RNNs (recurrent neural networks) to predict mutations in the SARS-CoV-2-selected non-structural proteins (NSP), i.e., NSP1, NSP3, NSP5, NSP8, NSP9, NSP13, and NSP15. First, we created and pre-processed the pairs of sequences from two languages using k-means clustering and nearest neighbors for training a neural translation machine. We also provided insights for training NMTs on long biological sequences. In addition, we evaluated and benchmarked our models to demonstrate their efficiency and reliability.

Concepts and methods for predicting viral evolution

The seasonal human influenza virus undergoes rapid evolution, leading to significant changes in circulating viral strains from year to year. These changes are typically driven by adaptive mutations, particularly in the antigenic epitopes, the regions of the viral surface protein haemagglutinin targeted by human antibodies. Here we describe a consistent set of methods for data-driven predictive analysis of viral evolution. Our pipeline integrates four types of data: (1) sequence data of viral isolates collected on a worldwide scale, (2) epidemiological data on incidences, (3) antigenic characterization of circulating viruses, and (4) intrinsic viral phenotypes. From the combined analysis of these data, we obtain estimates of relative fitness for circulating strains and predictions of clade frequencies for periods of up to one year. Furthermore, we obtain comparative estimates of protection against future viral populations for candidate vaccine strains, providing a basis for pre-emptive vaccine strain selection. Continuously updated predictions obtained from the prediction pipeline for influenza and SARS-CoV-2 are available on the website previr.app .

Concepts and methods for predicting viral evolution

The seasonal human influenza virus undergoes rapid evolution, leading to significant changes in circulating viral strains from year to year. These changes are typically driven by adaptive mutations, particularly in the antigenic epitopes, the regions of the viral surface protein haemagglutinin targeted by human antibodies. Here we describe a consistent set of methods for data-driven predictive analysis of viral evolution. Our pipeline integrates four types of data: (1) sequence data of viral isolates collected on a worldwide scale, (2) epidemiological data on incidences, (3) antigenic characterization of circulating viruses, and (4) intrinsic viral phenotypes. From the combined analysis of these data, we obtain estimates of relative fitness for circulating strains and predictions of clade frequencies for periods of up to one year. Furthermore, we obtain comparative estimates of protection against future viral populations for candidate vaccine strains, providing a basis for pre-emptive vaccine strain selection. Continuously updated predictions obtained from the prediction pipeline for influenza and SARS-CoV-2 are available on the website previr.app.

Emerging anti-spike monoclonal antibodies against SARS-CoV-2

INTRODUCTION

Anti-spike monoclonal antibodies (mAbs) were previously authorized for the prevention and treatment of COVID-19 in immunocompromised patients. However, they are no longer authorized in the U.S. due to their lack of neutralizing activity against current circulating SARS-CoV-2 Omicron variants.

AREAS COVERED

We summarized the available data on emergent mAbs in the early stages of clinical development. Consistent with data on prior mAbs, these novel agents have been well tolerated and demonstrated a good safety profile in early clinical trials. Additionally, many of them have been engineered to ensure prolonged half-life and combined with other mAbs to overcome the potential for emerging resistant mutants. Interestingly, one of these agents has been evaluated using an inhaled route of administration, and another agent is being evaluated for treatment of long COVID.

EXPERT OPINION

Although the available data of novel mAbs holds promise, we anticipate that these agents will face similar challenges encountered by prior authorized agents, including the continued evolution of SARS-CoV-2 and emergence of new escape mutations. Strategies to potentially mitigate this are discussed. Based on prior successful experience, immunocompromised patients will certainly benefit from the utilization of mAbs for the prevention and treatment of COVID-19; thus, we need to design potential interventions to ensure the sustained activity of these agents.

Antigenic drift expands viral escape pathways from imprinted host humoral immunity

An initial virus exposure can imprint antibodies such that future responses to antigenically drifted strains are dependent on the identity of the imprinting strain. Subsequent exposure to antigenically distinct strains followed by affinity maturation can guide immune responses toward generation of cross-reactive antibodies. How viruses evolve in turn to escape these imprinted broad antibody responses is unclear. Here, we used clonal antibody lineages from two human donors recognizing conserved influenza virus hemagglutinin (HA) epitopes to assess viral escape potential using deep mutational scanning. We show that even though antibody affinity maturation does restrict the number of potential escape routes in the imprinting strain through repositioning the antibody variable domains, escape is still readily observed in drifted strains and attributed to epistatic networks within HA. These data explain how influenza virus continues to evolve in the human population by escaping even broad antibody responses.

Innovative applications of artificial intelligence during the COVID-19 pandemic

The COVID-19 pandemic has created unprecedented challenges worldwide. Artificial intelligence (AI) technologies hold tremendous potential for tackling key aspects of pandemic management and response. In the present review, we discuss the tremendous possibilities of AI technology in addressing the global challenges posed by the COVID-19 pandemic. First, we outline the multiple impacts of the current pandemic on public health, the economy, and society. Next, we focus on the innovative applications of advanced AI technologies in key areas such as COVID-19 prediction, detection, control, and drug discovery for treatment. Specifically, AI-based predictive analytics models can use clinical, epidemiological, and omics data to forecast disease spread and patient outcomes. Additionally, deep neural networks enable rapid diagnosis through medical imaging. Intelligent systems can support risk assessment, decision-making, and social sensing, thereby improving epidemic control and public health policies. Furthermore, high-throughput virtual screening enables AI to accelerate the identification of therapeutic drug candidates and opportunities for drug repurposing. Finally, we discuss future research directions for AI technology in combating COVID-19, emphasizing the importance of interdisciplinary collaboration. Though promising, barriers related to model generalization, data quality, infrastructure readiness, and ethical risks must be addressed to fully translate these innovations into real-world impacts. Multidisciplinary collaboration engaging diverse expertise and stakeholders is imperative for developing robust, responsible, and human-centered AI solutions against COVID-19 and future public health emergencies.

Evaluating generalizability of artificial intelligence models for molecular datasets

Deep learning has made rapid advances in modeling molecular sequencing data. Despite achieving high performance on benchmarks, it remains unclear to what extent deep learning models learn general principles and generalize to previously unseen sequences. Benchmarks traditionally interrogate model generalizability by generating metadata based (MB) or sequence-similarity based (SB) train and test splits of input data before assessing model performance. Here, we show that this approach mischaracterizes model generalizability by failing to consider the full spectrum of cross-split overlap, i.e., similarity between train and test splits. We introduce Spectra, a spectral framework for comprehensive model evaluation. For a given model and input data, Spectra plots model performance as a function of decreasing cross-split overlap and reports the area under this curve as a measure of generalizability. We apply Spectra to 18 sequencing datasets with associated phenotypes ranging from antibiotic resistance in tuberculosis to protein-ligand binding to evaluate the generalizability of 19 state-of-the-art deep learning models, including large language models, graph neural networks, diffusion models, and convolutional neural networks. We show that SB and MB splits provide an incomplete assessment of model generalizability. With Spectra, we find as cross-split overlap decreases, deep learning models consistently exhibit a reduction in performance in a task- and model-dependent manner. Although no model consistently achieved the highest performance across all tasks, we show that deep learning models can generalize to previously unseen sequences on specific tasks. Spectra paves the way toward a better understanding of how foundation models generalize in biology.

Machine learning for functional protein design

Recent breakthroughs in AI coupled with the rapid accumulation of protein sequence and structure data have radically transformed computational protein design. New methods promise to escape the constraints of natural and laboratory evolution, accelerating the generation of proteins for applications in biotechnology and medicine. To make sense of the exploding diversity of machine learning approaches, we introduce a unifying framework that classifies models on the basis of their use of three core data modalities: sequences, structures and functional labels. We discuss the new capabilities and outstanding challenges for the practical design of enzymes, antibodies, vaccines, nanomachines and more. We then highlight trends shaping the future of this field, from large-scale assays to more robust benchmarks, multimodal foundation models, enhanced sampling strategies and laboratory automation.

ProteinNPT: Improving Protein Property Prediction and Design with Non-Parametric Transformers

Protein design holds immense potential for optimizing naturally occurring proteins, with broad applications in drug discovery, material design, and sustainability. However, computational methods for protein engineering are confronted with significant challenges, such as an expansive design space, sparse functional regions, and a scarcity of available labels. These issues are further exacerbated in practice by the fact most real-life design scenarios necessitate the simultaneous optimization of multiple properties. In this work, we introduce ProteinNPT, a non-parametric transformer variant tailored to protein sequences and particularly suited to label-scarce and multi-task learning settings. We first focus on the supervised fitness prediction setting and develop several cross-validation schemes which support robust performance assessment. We subsequently reimplement prior top-performing baselines, introduce several extensions of these baselines by integrating diverse branches of the protein engineering literature, and demonstrate that ProteinNPT consistently outperforms all of them across a diverse set of protein property prediction tasks. Finally, we demonstrate the value of our approach for iterative protein design across extensive in silico Bayesian optimization and conditional sampling experiments.

Full-spike deep mutational scanning helps predict the evolutionary success of SARS-CoV-2 clades

SARS-CoV-2 variants acquire mutations in spike that promote immune evasion and impact other properties that contribute to viral fitness such as ACE2 receptor binding and cell entry. Knowledge of how mutations affect these spike phenotypes can provide insight into the current and potential future evolution of the virus. Here we use pseudovirus deep mutational scanning to measure how >9,000 mutations across the full XBB.1.5 and BA.2 spikes affect ACE2 binding, cell entry, or escape from human sera. We find that mutations outside the receptor-binding domain (RBD) have meaningfully impacted ACE2 binding during SARS-CoV-2 evolution. We also measure how mutations to the XBB.1.5 spike affect neutralization by serum from individuals who recently had SARS-CoV-2 infections. The strongest serum escape mutations are in the RBD at sites 357, 420, 440, 456, and 473-however, the antigenic impacts of these mutations vary across individuals. We also identify strong escape mutations outside the RBD; however many of them decrease ACE2 binding, suggesting they act by modulating RBD conformation. Notably, the growth rates of human SARS-CoV-2 clades can be explained in substantial part by the measured effects of mutations on spike phenotypes, suggesting our data could enable better prediction of viral evolution.

Editorial: Factors Driving New Variants of SARS-CoV-2, Immune Escape, and Resistance to Antiviral Treatments as the End of the COVID-19 Pandemic is Declared

On May 5, 2023, the Director-General of the World Health Organization (WHO) advised the transition to long-term management of the COVID-19 pandemic and that COVID-19 is now an established and ongoing health issue that is no longer a public health emergency of international concern (PHEIC). The WHO decision was based on an analysis of the decreasing trend in mortality, the decline in hospital admissions and Intensive Care Unit (ICU) admissions from COVID-19, and the increasing levels of population immunity to SARS-CoV-2. This Editorial aims to highlight what is known of the factors that drive new variants, subvariants, and lineages of SARS-CoV-2 associated with immune escape to previous infection or vaccines and resistance to antiviral treatments as the end of the COVID-19 pandemic is declared.