Deep Mutational Scanning Data Research Articles

Flattening the curve-How to get better results with small deep-mutational-scanning datasets.

Proteins are used in various biotechnological applications, often requiring the optimization of protein properties by introducing specific amino-acid exchanges. Deep mutational scanning (DMS) is an effective high-throughput method for evaluating the effects of these exchanges on protein function. DMS data can then inform the training of a neural network to predict the impact of mutations. Most approaches use some representation of the protein sequence for training and prediction. As proteins are characterized by complex structures and intricate residue interaction networks, directly providing structural information as input reduces the need to learn these features from the data. We introduce a method for encoding protein structures as stacked 2D contact maps, which capture residue interactions, their evolutionary conservation, and mutation-induced interaction changes. Furthermore, we explored techniques to augment neural network training performance on smaller DMS datasets. To validate our approach, we trained three neural network architectures originally used for image analysis on three DMS datasets, and we compared their performances with networks trained solely on protein sequences. The results confirm the effectiveness of the protein structure encoding in machine learning efforts on DMS data. Using structural representations as direct input to the networks, along with data augmentation and pretraining, significantly reduced demands on training data size and improved prediction performance, especially on smaller datasets, while performance on large datasets was on par with state-of-the-art sequence convolutional neural networks. The methods presented here have the potential to provide the same workflow as DMS without the experimental and financial burden of testing thousands of mutants. Additionally, we present an open-source, user-friendly software tool to make these data analysis techniques accessible, particularly to biotechnology and protein engineering researchers who wish to apply them to their mutagenesis data.

CADD v1.7: using protein language models, regulatory CNNs and other nucleotide-level scores to improve genome-wide variant predictions.

Machine Learning-based scoring and classification of genetic variants aids the assessment of clinical findings and is employed to prioritize variants in diverse genetic studies and analyses. Combined Annotation-Dependent Depletion (CADD) is one of the first methods for the genome-wide prioritization of variants across different molecular functions and has been continuously developed and improved since its original publication. Here, we present our most recent release, CADD v1.7. We explored and integrated new annotation features, among them state-of-the-art protein language model scores (Meta ESM-1v), regulatory variant effect predictions (from sequence-based convolutional neural networks) and sequence conservation scores (Zoonomia). We evaluated the new version on data sets derived from ClinVar, ExAC/gnomAD and 1000 Genomes variants. For coding effects, we tested CADD on 31 Deep Mutational Scanning (DMS) data sets from ProteinGym and, for regulatory effect prediction, we used saturation mutagenesis reporter assay data of promoter and enhancer sequences. The inclusion of new features further improved the overall performance of CADD. As with previous releases, all data sets, genome-wide CADD v1.7 scores, scripts for on-site scoring and an easy-to-use webserver are readily provided via https://cadd.bihealth.org/ or https://cadd.gs.washington.edu/ to the community.

In silico prediction of immune-escaping hot spots for future COVID-19 vaccine design

The COVID-19 pandemic has had a widespread impact on a global scale, and the evolution of considerable dominants has already taken place. Some variants contained certain key mutations located on the receptor binding domain (RBD) of spike protein, such as E484K and N501Y. It is increasingly worrying that these variants could impair the efficacy of current vaccines or therapies. Therefore, analyzing and predicting the high-risk mutations of SARS-CoV-2 spike glycoprotein is crucial to design future vaccines against the different variants. In this work, we proposed an in silico approach, immune-escaping score (IES), to predict high-risk immune-escaping hot spots on the receptor-binding domain (RBD), implemented through integrated delta binding free energy measured by computational mutagenesis of spike-antibody complexes and mutation frequency calculated from viral genome sequencing data. We identified 23 potentially immune-escaping mutations on the RBD by using IES, nine of which occurred in omicron variants (R346K, K417N, N440K, L452Q, L452R, S477N, T478K, F490S, and N501Y), despite our dataset being curated before the omicron first appeared. The highest immune-escaping score (IES = 1) was found for E484K, which agrees with recent studies stating that the mutation significantly reduced the efficacy of neutralization antibodies. Furthermore, our predicted delta binding free energy and IES show a high correlation with high-throughput deep mutational scanning data (Pearson’s r = 0.70) and experimentally measured neutralization titers data (mean Pearson’s r = −0.80). In summary, our work presents a new method to identify the potentially immune-escaping mutations on the RBD and provides valuable insights into future COVID-19 vaccine design.

Cross-protein transfer learning substantially improves disease variant prediction

BackgroundGenetic variation in the human genome is a major determinant of individual disease risk, but the vast majority of missense variants have unknown etiological effects. Here, we present a robust learning framework for leveraging saturation mutagenesis experiments to construct accurate computational predictors of proteome-wide missense variant pathogenicity.ResultsWe train cross-protein transfer (CPT) models using deep mutational scanning (DMS) data from only five proteins and achieve state-of-the-art performance on clinical variant interpretation for unseen proteins across the human proteome. We also improve predictive accuracy on DMS data from held-out proteins. High sensitivity is crucial for clinical applications and our model CPT-1 particularly excels in this regime. For instance, at 95% sensitivity of detecting human disease variants annotated in ClinVar, CPT-1 improves specificity to 68%, from 27% for ESM-1v and 55% for EVE. Furthermore, for genes not used to train REVEL, a supervised method widely used by clinicians, we show that CPT-1 compares favorably with REVEL. Our framework combines predictive features derived from general protein sequence models, vertebrate sequence alignments, and AlphaFold structures, and it is adaptable to the future inclusion of other sources of information. We find that vertebrate alignments, albeit rather shallow with only 100 genomes, provide a strong signal for variant pathogenicity prediction that is complementary to recent deep learning-based models trained on massive amounts of protein sequence data. We release predictions for all possible missense variants in 90% of human genes.ConclusionsOur results demonstrate the utility of mutational scanning data for learning properties of variants that transfer to unseen proteins.

Learning protein fitness landscapes with deep mutational scanning data from multiple sources.

One of the key points of machine learning-assisted directed evolution (MLDE) is the accurate learning of the fitness landscape, a conceptual mapping from sequence variants to the desired function. Here, we describe a multi-protein training scheme that leverages the existing deep mutational scanning data from diverse proteins to aid in understanding the fitness landscape of a new protein. Proof-of-concept trials are designed to validate this training scheme in three aspects: random and positional extrapolation for single-variant effects, zero-shot fitness predictions for new proteins, and extrapolation for higher-order variant effects from single-variant effects. Moreover, our study identified previously overlooked strong baselines, and their unexpectedly good performance brings our attention to the pitfalls of MLDE. Overall, these results may improve our understanding of the association between different protein fitness profiles and shed light on developing better machine learning-assisted approaches to the directed evolution of proteins. A record of this paper's transparent peer review process is included in the supplemental information.

Allosteric Regulatory Control in Dihydrofolate Reductase is Revealed by Dynamic Asymmetry.

We investigated the relationship between mutations and dynamics in Escherichia coli dihydrofolate reductase (DHFR) using computational methods. Our study focused on the M20 and FG loops, which are known to be functionally important and affected by mutations distal to the loops. We used Molecular Dynamics simulations and developed position-specific metrics, including the Dynamic Flexibility Index (DFI) and Dynamic Coupling Index (DCI), to analyze the dynamics of wild-type DHFR and compared our results with existing deep mutational scanning data. Our analysis showed a statistically significant association between DFI and mutational tolerance of the DHFR positions, indicating that DFI can predict functionally beneficial or detrimental substitutions. We also applied an asymmetric version of our DCI metric (DCIasym ) to DHFR and found that certain distal residues control the dynamics of the M20 and FG loops, whereas others are controlled by them. Residues that are suggested to control the M20 and FG loops by our DCIasym metric are evolutionarily non-conserved; mutations at these sites can enhance enzyme activity. On the other hand, residues controlled by the loops are mostly deleterious to function when mutated and are also evolutionary conserved. Our results suggest that dynamics-based metrics can identify residues that explain the relationship between mutation and protein function or can be targeted to rationally engineer enzymes with enhanced activity. This article is protected by copyright. All rights reserved.

Updated benchmarking of variant effect predictors using deep mutational scanning.

The assessment of variant effect predictor (VEP) performance is fraught with biases introduced by benchmarking against clinical observations. In this study, building on our previous work, we use independently generated measurements of protein function from deep mutational scanning (DMS) experiments for 26 human proteins to benchmark 55 different VEPs, while introducing minimal data circularity. Many top-performing VEPs are unsupervised methods including EVE, DeepSequence and ESM-1v, a protein language model that ranked first overall. However, the strong performance of recent supervised VEPs, in particular VARITY, shows that developers are taking data circularity and bias issues seriously. We also assess the performance of DMS and unsupervised VEPs for discriminating between known pathogenic and putatively benign missense variants. Our findings are mixed, demonstrating that some DMS datasets perform exceptionally at variant classification, while others are poor. Notably, we observe a striking correlation between VEP agreement with DMS data and performance in identifying clinically relevant variants, strongly supporting the validity of our rankings and the utility of DMS for independent benchmarking.

Supplemental data files in "popDMS infers mutation effects from deep mutational scanning data"

Supplemental data files in "popDMS infers mutation effects from deep mutational scanning data"

Integrating deep mutational scanning and low-throughput mutagenesis data to predict the impact of amino acid variants.

Evaluating the impact of amino acid variants has been a critical challenge for studying protein function and interpreting genomic data. High-throughput experimental methods like deep mutational scanning (DMS) can measure the effect of large numbers of variants in a target protein, but because DMS studies have not been performed on all proteins, researchers also model DMS data computationally to estimate variant impacts by predictors. In this study, we extended a linear regression-based predictor to explore whether incorporating data from alanine scanning (AS), a widely used low-throughput mutagenesis method, would improve prediction results. To evaluate our model, we collected 146 AS datasets, mapping to 54 DMS datasets across 22 distinct proteins. We show that improved model performance depends on the compatibility of the DMS and AS assays, and the scale of improvement is closely related to the correlation between DMS and AS results.

A biophysical model of viral escape from polyclonal antibodies.

A challenge in studying viral immune escape is determining how mutations combine to escape polyclonal antibodies, which can potentially target multiple distinct viral epitopes. Here we introduce a biophysical model of this process that partitions the total polyclonal antibody activity by epitope and then quantifies how each viral mutation affects the antibody activity against each epitope. We develop software that can use deep mutational scanning data to infer these properties for polyclonal antibody mixtures. We validate this software using a computationally simulated deep mutational scanning experiment and demonstrate that it enables the prediction of escape by arbitrary combinations of mutations. The software described in this paper is available at https://jbloomlab.github.io/polyclonal.

Towards generalizable predictions for G protein-coupled receptor variant expression

Missense mutations that compromise the plasma membrane expression (PME) of integral membrane proteins are the root cause of numerous genetic diseases. Differentiation of this class of mutations from those that specifically modify the activity of the folded protein has proven useful for the development and targeting of precision therapeutics. Nevertheless, it remains challenging to predict the effects of mutations on the stability and/ or expression of membrane proteins. In this work, we utilize deep mutational scanning data to train a series of artificial neural networks to predict the PME of transmembrane domain variants of G protein-coupled receptors from structural and/ or evolutionary features. We show that our best-performing network, which we term the PME predictor, can recapitulate mutagenic trends within rhodopsin and can differentiate pathogenic transmembrane domain variants that cause it to misfold from those that compromise its signaling. This network also generates statistically significant predictions for the relative PME of transmembrane domain variants for another class A G protein-coupled receptor (β2 adrenergic receptor) but not for an unrelated voltage-gated potassium channel (KCNQ1). Notably, our analyses of these networks suggest structural features alone are generally sufficient to recapitulate the observed mutagenic trends. Moreover, our findings imply that networks trained in this manner may be generalizable to proteins that share a common fold. Implications of our findings for the design of mechanistically specific genetic predictors are discussed.

An antibody-escape estimator for mutations to the SARS-CoV-2 receptor-binding domain.

A key goal of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) surveillance is to rapidly identify viral variants with mutations that reduce neutralization by polyclonal antibodies elicited by vaccination or infection. Unfortunately, direct experimental characterization of new viral variants lags their sequence-based identification. Here we help address this challenge by aggregating deep mutational scanning data into an ‘escape estimator’ that estimates the antigenic effects of arbitrary combinations of mutations to the virus’s spike receptor-binding domain. The estimator can be used to intuitively visualize how mutations impact polyclonal antibody recognition and score the expected antigenic effect of combinations of mutations. These scores correlate with neutralization assays performed on SARS-CoV-2 variants and emphasize the ominous antigenic properties of the recently described Omicron variant. An interactive version of the estimator is at https://jbloomlab.github.io/SARS2_RBD_Ab_escape_maps/escape-calc/ (last accessed 11 March 2022), and we provide a Python module for batch processing. Currently the calculator uses primarily data for antibodies elicited by Wuhan-Hu-1-like vaccination or infection and so is expected to work best for calculating escape from such immunity for mutations relative to early SARS-CoV-2 strains.

Embeddings from protein language models predict conservation and variant effects

The emergence of SARS-CoV-2 variants stressed the demand for tools allowing to interpret the effect of single amino acid variants (SAVs) on protein function. While Deep Mutational Scanning (DMS) sets continue to expand our understanding of the mutational landscape of single proteins, the results continue to challenge analyses. Protein Language Models (pLMs) use the latest deep learning (DL) algorithms to leverage growing databases of protein sequences. These methods learn to predict missing or masked amino acids from the context of entire sequence regions. Here, we used pLM representations (embeddings) to predict sequence conservation and SAV effects without multiple sequence alignments (MSAs). Embeddings alone predicted residue conservation almost as accurately from single sequences as ConSeq using MSAs (two-state Matthews Correlation Coefficient—MCC—for ProtT5 embeddings of 0.596 ± 0.006 vs. 0.608 ± 0.006 for ConSeq). Inputting the conservation prediction along with BLOSUM62 substitution scores and pLM mask reconstruction probabilities into a simplistic logistic regression (LR) ensemble for Variant Effect Score Prediction without Alignments (VESPA) predicted SAV effect magnitude without any optimization on DMS data. Comparing predictions for a standard set of 39 DMS experiments to other methods (incl. ESM-1v, DeepSequence, and GEMME) revealed our approach as competitive with the state-of-the-art (SOTA) methods using MSA input. No method outperformed all others, neither consistently nor statistically significantly, independently of the performance measure applied (Spearman and Pearson correlation). Finally, we investigated binary effect predictions on DMS experiments for four human proteins. Overall, embedding-based methods have become competitive with methods relying on MSAs for SAV effect prediction at a fraction of the costs in computing/energy. Our method predicted SAV effects for the entire human proteome (~ 20 k proteins) within 40 min on one Nvidia Quadro RTX 8000. All methods and data sets are freely available for local and online execution through bioembeddings.com, https://github.com/Rostlab/VESPA, and PredictProtein.

Neural networks to learn protein sequence–function relationships from deep mutational scanning data

The mapping from protein sequence to function is highly complex, making it challenging to predict how sequence changes will affect a protein's behavior and properties. We present a supervised deep learning framework to learn the sequence-function mapping from deep mutational scanning data and make predictions for new, uncharacterized sequence variants. We test multiple neural network architectures, including a graph convolutional network that incorporates protein structure, to explore how a network's internal representation affects its ability to learn the sequence-function mapping. Our supervised learning approach displays superior performance over physics-based and unsupervised prediction methods. We find that networks that capture nonlinear interactions and share parameters across sequence positions are important for learning the relationship between sequence and function. Further analysis of the trained models reveals the networks' ability to learn biologically meaningful information about protein structure and mechanism. Finally, we demonstrate the models' ability to navigate sequence space and design new proteins beyond the training set. We applied the proteinG B1 domain (GB1) models to design a sequence that binds to immunoglobulin G with substantially higher affinity than wild-type GB1.

Stabilization of the SARS-CoV-2 Spike Receptor-Binding Domain Using Deep Mutational Scanning and Structure-Based Design.

The unprecedented global demand for SARS-CoV-2 vaccines has demonstrated the need for highly effective vaccine candidates that are thermostable and amenable to large-scale manufacturing. Nanoparticle immunogens presenting the receptor-binding domain (RBD) of the SARS-CoV-2 Spike protein (S) in repetitive arrays are being advanced as second-generation vaccine candidates, as they feature robust manufacturing characteristics and have shown promising immunogenicity in preclinical models. Here, we used previously reported deep mutational scanning (DMS) data to guide the design of stabilized variants of the RBD. The selected mutations fill a cavity in the RBD that has been identified as a linoleic acid binding pocket. Screening of several designs led to the selection of two lead candidates that expressed at higher yields than the wild-type RBD. These stabilized RBDs possess enhanced thermal stability and resistance to aggregation, particularly when incorporated into an icosahedral nanoparticle immunogen that maintained its integrity and antigenicity for 28 days at 35-40°C, while corresponding immunogens displaying the wild-type RBD experienced aggregation and loss of antigenicity. The stabilized immunogens preserved the potent immunogenicity of the original nanoparticle immunogen, which is currently being evaluated in a Phase I/II clinical trial. Our findings may improve the scalability and stability of RBD-based coronavirus vaccines in any format and more generally highlight the utility of comprehensive DMS data in guiding vaccine design.

Global Dynamics Renders Protein Sites with High Functional Response.

Deep mutational scanning enables examination of the effects of many mutations at each amino acid position in a query protein, readily disclosing positions that are particularly sensitive. Mutations in these positions alter protein function the most. Here, on the premise that dynamics underlie function, we explore to what extent the measured sensitivity to mutations could be linked to-perhaps be explained by-the structural dynamics of the protein. We employ a minimalist perturbation-response approach based on the Gaussian Network Model (GNM) on a data set of seven proteins with deep mutational scanning data. The analysis shows that the mutation-sensitive positions are often of capacity to modulate the global dynamics and to intermediate allosteric interactions in the structure. With that, upon strain perturbation, these positions decrease residue fluctuations the most, affecting function via entropy changes. This is particularly relevant for positions that are distant from binding sites or other functional regions of the protein and are sensitive to mutations, nevertheless. Our results indicate that mutations in these positions allosterically manipulate protein function.

Inferring Protein Sequence-Function Relationships with Large-Scale Positive-Unlabeled Learning.

Machine learning can infer how protein sequence maps to function without requiring a detailed understanding of the underlying physical or biological mechanisms. It is challenging to apply existing supervised learning frameworks to large-scale experimental data generated by deep mutational scanning (DMS) and related methods. DMS data often contain high-dimensional and correlated sequence variables, experimental sampling error and bias, and the presence of missing data. Notably, most DMS data do not contain examples of negative sequences, making it challenging to directly estimate how sequence affects function. Here, we develop a positive-unlabeled (PU) learning framework to infer sequence-function relationships from large-scale DMS data. Our PU learning method displays excellent predictive performance across ten large-scale sequence-function datasets, representing proteins of different folds, functions, and library types. The estimated parameters pinpoint key residues that dictate protein structure and function. Finally, we apply our statistical sequence-function model to design highly stabilized enzymes.

Toward Developing Intuitive Rules for Protein Variant Effect Prediction Using Deep Mutational Scanning Data.

Protein structure and function can be severely altered by even a single amino acid mutation. Predictions of mutational effects using extensive artificial intelligence (AI)-based models, although accurate, remain as enigmatic as the experimental observations in terms of improving intuitions about the contributions of various factors. Inspired by Lipinski’s rules for drug-likeness, we devise simple thresholding criteria on five different descriptors such as conservation, which have so far been limited to qualitative interpretations such as high conservation implies high mutational effect. We analyze systematic deep mutational scanning data of all possible single amino acid substitutions on seven proteins (25153 mutations) to first define these thresholds and then to evaluate the scope and limits of the predictions. At this stage, the approach allows us to comment easily and with a low error rate on the subset of mutations classified as neutral or deleterious by all of the descriptors. We hope that complementary to the accurate AI predictions, these thresholding rules or their subsequent modifications will serve the purpose of codifying the knowledge about the effects of mutations.

Dms-view: Interactive visualization tool for deep mutational scanning data.

The high-throughput technique of deep mutational scanning (DMS) has recently made it possible to experimentally measure the effects of all amino-acid mutations to a protein. Over the past five years, this technique has been used to study dozens of different proteins and answer a variety of research questions. For example, DMS has been used for protein engineering, understanding the human immune response to viruses, and interpreting human variation in a clinical setting. Accompanying this proliferation of DMS studies has been the development of software tools and databases for data analysis and sharing. However, for many purposes it is important to also integrate and visualize the DMS data in the context of other information, such as the 3-D protein structure or natural sequence-variation data. Here we describe dms-view (https://dms-view.github.io/), a flexible, web-based, interactive visualization tool for DMS data. dms-view is written in JavaScript and D3, and links site-level and mutation-level DMS data to a 3-D protein structure. The user can interactively select sites of interest to examine the DMS measurements in the context of the protein structure. dms-view tracks the input data and user selections in the URL, making it possible to save specific views of interactively generated visualizations to share with collaborators or to support a published study. Importantly, dms-view takes a flexible input data file so users can easily visualize their own DMS data in the context of protein structures of their choosing, and also incorporate additional information such amino-acid frequencies in natural alignments.

DiMSum: an error model and pipeline for analyzing deep mutational scanning data and diagnosing common experimental pathologies

Deep mutational scanning (DMS) enables multiplexed measurement of the effects of thousands of variants of proteins, RNAs, and regulatory elements. Here, we present a customizable pipeline, DiMSum, that represents an end-to-end solution for obtaining variant fitness and error estimates from raw sequencing data. A key innovation of DiMSum is the use of an interpretable error model that captures the main sources of variability arising in DMS workflows, outperforming previous methods. DiMSum is available as an R/Bioconda package and provides summary reports to help researchers diagnose common DMS pathologies and take remedial steps in their analyses.