High-resolution mapping of protein sequence-function relationships

Integration of FRET and sequencing to engineer kinase biosensors from mammalian cell libraries

The limited sensitivity of Förster Resonance Energy Transfer (FRET) biosensors hinders their broader applications. Here, we develop an approach integrating high-throughput FRET sorting and next-generation sequencing (FRET-Seq) to identify sensitive biosensors with varying substrate sequences from large-scale libraries directly in mammalian cells, utilizing the design of self-activating FRET (saFRET) biosensor. The resulting biosensors of Fyn and ZAP70 kinases exhibit enhanced performance and enable the dynamic imaging of T-cell activation mediated by T cell receptor (TCR) or chimeric antigen receptor (CAR), revealing a highly organized ZAP70 subcellular activity pattern upon TCR but not CAR engagement. The ZAP70 biosensor elucidates the role of immunoreceptor tyrosine-based activation motif (ITAM) in affecting ZAP70 activation to regulate CAR functions. A saFRET biosensor-based high-throughput drug screening (saFRET-HTDS) assay further enables the identification of an FDA-approved cancer drug, Sunitinib, that can be repurposed to inhibit ZAP70 activity and autoimmune-disease-related T-cell activation.

Revealing enzyme functional architecture via high-throughput microfluidic enzyme kinetics

Systematic and extensive investigation of enzymes is needed to understand their extraordinary efficiency and meet current challenges in medicine and engineering. We present HT-MEK (High-Throughput Microfluidic Enzyme Kinetics), a microfluidic platform for high-throughput expression, purification, and characterization of more than 1500 enzyme variants per experiment. For 1036 mutants of the alkaline phosphatase PafA (phosphate-irrepressible alkaline phosphatase of Flavobacterium), we performed more than 670,000 reactions and determined more than 5000 kinetic and physical constants for multiple substrates and inhibitors. We uncovered extensive kinetic partitioning to a misfolded state and isolated catalytic effects, revealing spatially contiguous regions of residues linked to particular aspects of function. Regions included active-site proximal residues but extended to the enzyme surface, providing a map of underlying architecture not possible to derive from existing approaches. HT-MEK has applications that range from understanding molecular mechanisms to medicine, engineering, and design.

Systems Approaches to Unravel T Cell Function and Therapeutic Potential in Autoimmune Disease

Producing Ag-specific immune responses constrained to target tissues or cells that can be engaged or disengaged at will is predicated on understanding the network of genes governing immune cell function, defining the rules underlying Ag specificity, and synthesizing the tools to engineer them. The successes and limitations of chimeric Ag receptor (CAR) T cells emphasize this goal, and advances in high-throughput sequencing, large-scale genomic screens, single-cell profiling, and genetic modification are providing the necessary data to bring it to fruition-including a broader application into the treatment of autoimmune diseases. In this review, we delve into the implementation of these developments, survey the relevant works, and propose a framework for generating the next generation of synthetic T cells informed by the principles learned from these systems approaches.

Evotuning protocols for Transformer-based variant effect prediction on multi-domain proteins

Accurate variant effect prediction has broad impacts on protein engineering. Recent machine learning approaches toward this end are based on representation learning, by which feature vectors are learned and generated from unlabeled sequences. However, it is unclear how to effectively learn evolutionary properties of an engineering target protein from homologous sequences, taking into account the protein's sequence-level structure called domain architecture (DA). Additionally, no optimal protocols are established for incorporating such properties into Transformer, the neural network well-known to perform the best in natural language processing research. This article proposes DA-aware evolutionary fine-tuning, or 'evotuning', protocols for Transformer-based variant effect prediction, considering various combinations of homology search, fine-tuning and sequence vectorization strategies. We exhaustively evaluated our protocols on diverse proteins with different functions and DAs. The results indicated that our protocols achieved significantly better performances than previous DA-unaware ones. The visualizations of attention maps suggested that the structural information was incorporated by evotuning without direct supervision, possibly leading to better prediction accuracy.

The Adaptive Potential of the Middle Domain of Yeast Hsp90

The distribution of fitness effects (DFEs) of new mutations across different environments quantifies the potential for adaptation in a given environment and its cost in others. So far, results regarding the cost of adaptation across environments have been mixed, and most studies have sampled random mutations across different genes. Here, we quantify systematically how costs of adaptation vary along a large stretch of protein sequence by studying the distribution of fitness effects of the same ≈2,300 amino-acid changing mutations obtained from deep mutational scanning of 119 amino acids in the middle domain of the heat shock protein Hsp90 in five environments. This region is known to be important for client binding, stabilization of the Hsp90 dimer, stabilization of the N-terminal-Middle and Middle-C-terminal interdomains, and regulation of ATPase–chaperone activity. Interestingly, we find that fitness correlates well across diverse stressful environments, with the exception of one environment, diamide. Consistent with this result, we find little cost of adaptation; on average only one in seven beneficial mutations is deleterious in another environment. We identify a hotspot of beneficial mutations in a region of the protein that is located within an allosteric center. The identified protein regions that are enriched in beneficial, deleterious, and costly mutations coincide with residues that are involved in the stabilization of Hsp90 interdomains and stabilization of client-binding interfaces, or residues that are involved in ATPase–chaperone activity of Hsp90. Thus, our study yields information regarding the role and adaptive potential of a protein sequence that complements and extends known structural information.

Unsupervised Inference of Protein Fitness Landscape from Deep Mutational Scan

The recent technological advances underlying the screening of large combinatorial libraries in high-throughput mutational scans deepen our understanding of adaptive protein evolution and boost its applications in protein design. Nevertheless, the large number of possible genotypes requires suitable computational methods for data analysis, the prediction of mutational effects, and the generation of optimized sequences. We describe a computational method that, trained on sequencing samples from multiple rounds of a screening experiment, provides a model of the genotype-fitness relationship. We tested the method on five large-scale mutational scans, yielding accurate predictions of the mutational effects on fitness. The inferred fitness landscape is robust to experimental and sampling noise and exhibits high generalization power in terms of broader sequence space exploration and higher fitness variant predictions. We investigate the role of epistasis and show that the inferred model provides structural information about the 3D contacts in the molecular fold.

Optimization of therapeutic antibodies by predicting antigen specificity from antibody sequence via deep learning

The optimization of therapeutic antibodies is time-intensive and resource-demanding, largely because of the low-throughput screening of full-length antibodies (approximately 1 × 10³ variants) expressed in mammalian cells, which typically results in few optimized leads. Here we show that optimized antibody variants can be identified by predicting antigen specificity via deep learning from a massively diverse space of antibody sequences. To produce data for training deep neural networks, we deep-sequenced libraries of the therapeutic antibody trastuzumab (about 1 × 10⁴ variants), expressed in a mammalian cell line through site-directed mutagenesis via CRISPR-Cas9-mediated homology-directed repair, and screened the libraries for specificity to human epidermal growth factor receptor 2 (HER2). We then used the trained neural networks to screen a computational library of approximately 1 × 10⁸ trastuzumab variants and predict the HER2-specific subset (approximately 1 × 10⁶ variants), which can then be filtered for viscosity, clearance, solubility and immunogenicity to generate thousands of highly optimized lead candidates. Recombinant expression and experimental testing of 30 randomly selected variants from the unfiltered library showed that all 30 retained specificity for HER2. Deep learning may facilitate antibody engineering and optimization.

Computational-Driven Epitope Verification and Affinity Maturation of TLR4-Targeting Antibodies

Toll-like receptor (TLR) signaling plays a critical role in the induction and progression of autoimmune diseases such as rheumatoid arthritis, systemic lupus erythematous, experimental autoimmune encephalitis, type 1 diabetes mellitus and neurodegenerative diseases. Deciphering antigen recognition by antibodies provides insights and defines the mechanism of action into the progression of immune responses. Multiple strategies, including phage display and hybridoma technologies, have been used to enhance the affinity of antibodies for their respective epitopes. Here, we investigate the TLR4 antibody-binding epitope by computational-driven approach. We demonstrate that three important residues, i.e., Y328, N329, and K349 of TLR4 antibody binding epitope identified upon in silico mutagenesis, affect not only the interaction and binding affinity of antibody but also influence the structural integrity of TLR4. Furthermore, we predict a novel epitope at the TLR4-MD2 interface which can be targeted and explored for therapeutic antibodies and small molecules. This technique provides an in-depth insight into antibody-antigen interactions at the resolution and will be beneficial for the development of new monoclonal antibodies. Computational techniques, if coupled with experimental methods, will shorten the duration of rational design and development of antibody therapeutics.

UMAP as a Dimensionality Reduction Tool for Molecular Dynamics Simulations of Biomacromolecules: A Comparison Study

Proteins are the molecular machines of life. The multitude of possible conformations that proteins can adopt determines their free-energy landscapes. However, the inherently high dimensionality of a protein free-energy landscape poses a challenge to deciphering how proteins perform their functions. For this reason, dimensionality reduction is an active field of research for molecular biologists. The uniform manifold approximation and projection (UMAP) is a dimensionality reduction method based on a fuzzy topological analysis of data. In the present study, the performance of UMAP is compared with that of other popular dimensionality reduction methods such as t-distributed stochastic neighbor embedding (t-SNE), principal component analysis (PCA), and time-structure independent components analysis (tICA) in the context of analyzing molecular dynamics simulations of the circadian clock protein VIVID. A good dimensionality reduction method should accurately represent the data structure on the projected components. The comparison of the raw high-dimensional data with the projections obtained using different dimensionality reduction methods based on various metrics showed that UMAP has superior performance when compared with linear reduction methods (PCA and tICA) and has competitive performance and scalable computational cost.

Deep Mutational Scanning of Viral Glycoproteins and Their Host Receptors

Deep mutational scanning or deep mutagenesis is a powerful tool for understanding the sequence diversity available to viruses for adaptation in a laboratory setting. It generally involves tracking an in vitro selection of protein sequence variants with deep sequencing to map mutational effects based on changes in sequence abundance. Coupled with any of a number of selection strategies, deep mutagenesis can explore the mutational diversity available to viral glycoproteins, which mediate critical roles in cell entry and are exposed to the humoral arm of the host immune response. Mutational landscapes of viral glycoproteins for host cell attachment and membrane fusion reveal extensive epistasis and potential escape mutations to neutralizing antibodies or other therapeutics, as well as aiding in the design of optimized immunogens for eliciting broadly protective immunity. While less explored, deep mutational scans of host receptors further assist in understanding virus-host protein interactions. Critical residues on the host receptors for engaging with viral spikes are readily identified and may help with structural modeling. Furthermore, mutations may be found for engineering soluble decoy receptors as neutralizing agents that specifically bind viral targets with tight affinity and limited potential for viral escape. By untangling the complexities of how sequence contributes to viral glycoprotein and host receptor interactions, deep mutational scanning is impacting ideas and strategies at multiple levels for combatting circulating and emergent virus strains.

Parameters and determinants of responses to selection in antibody libraries

The sequences of antibodies from a given repertoire are highly diverse at few sites located on the surface of a genome-encoded larger scaffold. The scaffold is often considered to play a lesser role than highly diverse, non-genome-encoded sites in controlling binding affinity and specificity. To gauge the impact of the scaffold, we carried out quantitative phage display experiments where we compare the response to selection for binding to four different targets of three different antibody libraries based on distinct scaffolds but harboring the same diversity at randomized sites. We first show that the response to selection of an antibody library may be captured by two measurable parameters. Second, we provide evidence that one of these parameters is determined by the degree of affinity maturation of the scaffold, affinity maturation being the process by which antibodies accumulate somatic mutations to evolve towards higher affinities during the natural immune response. In all cases, we find that libraries of antibodies built around maturated scaffolds have a lower response to selection to other arbitrary targets than libraries built around germline-based scaffolds. We thus propose that germline-encoded scaffolds have a higher selective potential than maturated ones as a consequence of a selection for this potential over the long-term evolution of germline antibody genes. Our results are a first step towards quantifying the evolutionary potential of biomolecules.

Antibodies in the immune system consist of a genetically encoded scaffold that exposes a few highly diverse, non-genetically encoded sites. This focused diversity is sufficient to produce antibodies that bind to any target molecule. To understand the role of the scaffold, which acquires hypermutations during the immune response, over the selective response, we analyze quantitative in vitro experiments where large antibody populations based on different scaffolds are selected against different targets. We show that selective responses are described statistically by two parameters, one of which depends on prior evolution of the scaffold as part of a previous response. Our work provides methods to assay whether naïve antibody scaffolds are endowed with a distinctively high selective potential.

Disentangling the Contribution of Each Descriptive Characteristic of Every Single Mutation to Its Functional Effects

Mutational effects predictions continue to improve in accuracy as advanced artificial intelligence (AI) algorithms are trained on exhaustive experimental data. The next natural questions to ask are if it is possible to gain insights into which attribute of the mutation contributes how much to the mutational effects and if one can develop universal rules for mapping the descriptors to mutational effects. In this work, we mainly address the former aspect using a framework of interpretable AI. Relations between the physicochemical descriptors and their contributions to the mutational effects are extracted by analyzing the data on 29,832 variants from eight systematic deep mutational scan studies. An opposite trend in the dependence of fitness and solubility on the distance of the amino acid from the catalytic sites could be extracted and quantified. The dependence of the mutational effect contributions on the position-specific scoring matrix (PSSM) score for the amino acid after mutation or the BLOSUM score of the substitution showed universal trends. Our attempts in the present work to explain the quantitative differences in the dependence on conservation and SASA across proteins were not successful. The work nevertheless brings transparency into the predictions and development of rules, and will hopefully lead to empirically uncovering the universalities among these rules.

Local Interactions That Contribute Minimal Frustration Determine Foldability

Earlier experiments suggest that the evolutionary information (conservation and coevolution) encoded in protein sequences is necessary and sufficient to specify the fold of a protein family. However, there is no computational work to quantify the effect of such evolutionary information on the folding process. Here we explore the role of early folding steps for sequences designed using coevolution and conservation through a combination of computational and experimental methods. We simulated a repertoire of native and designed WW domain sequences to analyze early local contact formation and found that the N-terminal β-hairpin turn would not form correctly due to strong non-native local contacts in unfoldable sequences. Through a maximum likelihood approach, we identified five local contacts that play a critical role in folding, suggesting that a small subset of amino acid pairs can be used to solve the "needle in the haystack" problem to design foldable sequences. Thus, using the contact probability of those five local contacts that form during the early stage of folding, we built a classification model that predicts the foldability of a WW sequence with 81% accuracy. This classification model was used to redesign WW domain sequences that could not fold due to frustration and make them foldable by introducing a few mutations that led to the stabilization of these critical local contacts. The experimental analysis shows that a redesigned sequence folds and binds to polyproline peptides with a similar affinity as those observed for native WW domains. Overall, our analysis shows that evolutionary-designed sequences should not only satisfy the folding stability but also ensure a minimally frustrated folding landscape.

Prediction of Function Determining and Buried Residues Through Analysis of Saturation Mutagenesis Datasets

Mutational scanning can be used to probe effects of large numbers of point mutations on protein function. Positions affected by mutation are primarily at either buried or at exposed residues directly involved in function, hereafter designated as active-site residues. In the absence of prior structural information, it has not been easy to distinguish between these two categories of residues. We curated and analyzed a set of twelve published deep mutational scanning datasets. The analysis revealed differential patterns of mutational sensitivity and substitution preferences at buried and exposed positions. Prediction of buried-sites solely from the mutational sensitivity data was facilitated by incorporating predicted sequence-based accessibility values. For active-site residues we observed mean sensitivity, specificity and accuracy of 61, 90 and 88% respectively. For buried residues the corresponding figures were 59, 90 and 84% while for exposed non active-site residues these were 98, 44 and 82% respectively. We also identified positions which did not follow these general trends and might require further experimental re-validation. This analysis highlights the ability of deep mutational scans to provide important structural and functional insights, even in the absence of three-dimensional structures determined using conventional structure determination techniques, and also discuss some limitations of the methodology.

A large-scale survey of pairwise epistasis reveals a mechanism for evolutionary expansion and specialization of PDZ domains

Deep mutational scanning (DMS) facilitates data-driven models of protein structure and function. Here, we adapted Saturated Programmable Insertion Engineering (SPINE) as a programmable DMS technique. We validate SPINE with a reference single mutant dataset in the PSD95 PDZ3 domain and then characterize most pairwise double mutants to study epistasis. We observe wide-spread proximal negative epistasis, which we attribute to mutations affecting thermodynamic stability, and strong long-range positive epistasis, which is enriched in an evolutionarily conserved and function-defining network of "sector" and clade-specifying residues. Conditional neutrality of mutations in clade-specifying residues compensates for deleterious mutations in sector positions. This suggests that epistatic interactions between these position pairs facilitated the evolutionary expansion and specialization of PDZ domains. We propose that SPINE provides easy experimental access to reveal epistasis signatures in proteins that will improve our understanding of the structural basis for protein function and adaptation.

Modeling mutational effects on biochemical phenotypes using convolutional neural networks: application to SARS-CoV-2

Biochemical phenotypes are major indexes for protein structure and function characterization. They are determined, at least in part, by the intrinsic physicochemical properties of amino acids and may be reflected in the protein three-dimensional structure. Modeling mutational effects on biochemical phenotypes is a critical step for understanding protein function and disease mechanism as well as enabling drug discovery. Deep Mutational Scanning (DMS) experiments have been performed on SARS-CoV-2’s spike receptor binding domain and the human ACE2 zinc-binding peptidase domain - both central players in viral infection and evolution and antibody evasion - quantifying how mutations impact binding affinity and protein expression. Here, we modeled biochemical phenotypes from massively parallel assays, using convolutional neural networks trained on protein sequence mutations in the virus and human host. We found that neural networks are significantly predictive of binding affinity, protein expression, and antibody escape, learning complex interactions and higher-order features that are difficult to capture with conventional methods from structural biology. Integrating the intrinsic physicochemical properties of amino acids, including hydrophobicity, solvent-accessible surface area, and long-range non-bonded energy per atom, significantly improved prediction (empirical p<0.01) though there was such a strong dependence on the sequence data alone to yield reasonably good prediction. We observed concordance of the DMS data and our neural network predictions with an independent study on intermolecular interactions from molecular dynamics (multiple 500 ns or 1 μs all-atom) simulations of the spike protein-ACE2 interface, with critical implications for the use of deep learning to dissect molecular mechanisms. The mutation- or genetically-determined component of a biochemical phenotype estimated from the neural networks has improved causal inference properties relative to the original phenotype and can facilitate crucial insights into disease pathophysiology and therapeutic design.

SequencErr: measuring and suppressing sequencer errors in next-generation sequencing data

Background

There is currently no method to precisely measure the errors that occur in the sequencing instrument/sequencer, which is critical for next-generation sequencing applications aimed at discovering the genetic makeup of heterogeneous cellular populations.

Results

We propose a novel computational method, SequencErr, to address this challenge by measuring the base correspondence between overlapping regions in forward and reverse reads. An analysis of 3777 public datasets from 75 research institutions in 18 countries revealed the sequencer error rate to be ~ 10 per million (pm) and 1.4% of sequencers and 2.7% of flow cells have error rates > 100 pm. At the flow cell level, error rates are elevated in the bottom surfaces and > 90% of HiSeq and NovaSeq flow cells have at least one outlier error-prone tile. By sequencing a common DNA library on different sequencers, we demonstrate that sequencers with high error rates have reduced overall sequencing accuracy, and removal of outlier error-prone tiles improves sequencing accuracy. We demonstrate that SequencErr can reveal novel insights relative to the popular quality control method FastQC and achieve a 10-fold lower error rate than popular error correction methods including Lighter and Musket.

Conclusions

Our study reveals novel insights into the nature of DNA sequencing errors incurred on DNA sequencers. Our method can be used to assess, calibrate, and monitor sequencer accuracy, and to computationally suppress sequencer errors in existing datasets.

Exploring mutable conserved sites and fatal non-conserved sites by random mutation of esterase from Sulfolobus tokodaii and subtilisin from Thermococcus kodakarensis

Homologous proteins differ in their amino acid sequences at several positions. Generally, conserved sites are recognized as not suitable for amino acid substitution, and thus in evolutionary protein engineering, non-conserved sites are often selected as mutation sites. However, there have also been reports of possible mutations in conserved sites. In this study, we explored mutable conserved sites and immutable non-conserved sites by testing random mutations of two thermostable proteins, an esterase from Sulfolobus tokodaii (Sto-Est) and a subtilisin from Thermococcus kodakarensis (Tko-Sub). The subtilisin domain of Tko-Sub needs Ca²⁺ ions and the propeptide domain for stability, folding and maturation. The results from the two proteins showed that about one-third of the mutable sites were detected in conserved sites and some non-conserved sites lost enzymatic activity at high temperatures due to mutation. Of the conserved sites in Sto-Est, the sites on the loop, on the surface, and far from the active site are more resistant to mutation. In Tko-Sub, the sites flanking Ca²⁺-binding sites and propeptide were undesirable for mutation. The results presented here serve as an index for selecting mutation sites and contribute to the expansion of available sequence range by introducing mutations at conserved sites.

Multiplexed Functional Assessment of Genetic Variants in CARD11

Genetic testing has increased the number of variants identified in disease genes, but the diagnostic utility is limited by lack of understanding variant function. CARD11 encodes an adaptor protein that expresses dominant-negative and gain-of-function variants associated with distinct immunodeficiencies. Here, we used a "cloning-free" saturation genome editing approach in a diploid cell line to simultaneously score 2,542 variants for decreased or increased function in the region of CARD11 associated with immunodeficiency. We also described an exon-skipping mechanism for CARD11 dominant-negative activity. The classification of reported clinical variants was sensitive (94.6%) and specific (88.9%), which rendered the data immediately useful for interpretation of seven coding and splicing variants implicated in immunodeficiency found in our clinic. This approach is generalizable for variant interpretation in many other clinically actionable genes, in any relevant cell type.

UMI-linked consensus sequencing enables phylogenetic analysis of directed evolution

The success of protein evolution campaigns is strongly dependent on the sequence context in which mutations are introduced, stemming from pervasive non-additive interactions between a protein's amino acids ('intra-gene epistasis'). Our limited understanding of such epistasis hinders the correct prediction of the functional contributions and adaptive potential of mutations. Here we present a straightforward unique molecular identifier (UMI)-linked consensus sequencing workflow (UMIC-seq) that simplifies mapping of evolutionary trajectories based on full-length sequences. Attaching UMIs to gene variants allows accurate consensus generation for closely related genes with nanopore sequencing. We exemplify the utility of this approach by reconstructing the artificial phylogeny emerging in three rounds of directed evolution of an amine dehydrogenase biocatalyst via ultrahigh throughput droplet screening. Uniquely, we are able to identify lineages and their founding variant, as well as non-additive interactions between mutations within a full gene showing sign epistasis. Access to deep and accurate long reads will facilitate prediction of key beneficial mutations and adaptive potential based on in silico analysis of large sequence datasets.

New Strategies for Clinical Trials in Autism Spectrum Disorder

BACKGROUND

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition that poses several challenges in terms of clinical diagnosis and investigation of molecular etiology. The lack of knowledge on the pathogenic mechanisms underlying ASD has hampered the clinical trials that so far have tried to target ASD behavioral symptoms. In order to improve our understanding of the molecular abnormalities associated with ASD, a deeper and more extensive genetic profiling of targeted individuals with ASD was needed.

METHODS

The recent availability of new and more powerful sequencing technologies (third-generation sequencing) has allowed to develop novel strategies for the characterization of comprehensive genetic profiles of individuals with ASD. In particular, this review will describe integrated approaches based on the combination of various omics technologies that will lead to a better stratification of targeted cohorts for the design of clinical trials in ASD.

RESULTS

In order to analyze the big data collected by assays such as the whole genome, epigenome, transcriptome, and proteome, it is critical to develop an efficient computational infrastructure. Machine learning models are instrumental to identify non-linear relationships between the omics technologies and, therefore, establish a functional informative network among the different data sources.

CONCLUSION

The potential advantage provided by these new integrated omics-based strategies is better characterization of the genetic background of ASD cohorts, to identify novel molecular targets for drug development, and ultimately offer a more personalized approach in the design of clinical trials for ASD.

Uncovering the basis of protein-protein interaction specificity with a combinatorially complete library

Protein-protein interaction specificity is often encoded at the primary sequence level. However, the contributions of individual residues to specificity are usually poorly understood and often obscured by mutational robustness, sequence degeneracy, and epistasis. Using bacterial toxin-antitoxin systems as a model, we screened a combinatorially complete library of antitoxin variants at three key positions against two toxins. This library enabled us to measure the effect of individual substitutions on specificity in hundreds of genetic backgrounds. These distributions allow inferences about the general nature of interface residues in promoting specificity. We find that positive and negative contributions to specificity are neither inherently coupled nor mutually exclusive. Further, a wild-type antitoxin appears optimized for specificity as no substitutions improve discrimination between cognate and non-cognate partners. By comparing crystal structures of paralogous complexes, we provide a rationale for our observations. Collectively, this work provides a generalizable approach to understanding the logic of molecular recognition.

Functional analysis of BRCA1 RING domain variants: computationally derived structural data can improve upon experimental features for training predictive models

Advancements in the interpretation of variants of unknown significance are critical for improving clinical outcomes. In a recent study, massive parallel assays were used to experimentally quantify the effects of missense substitutions in the RING domain of BRCA1 on E3 ubiquitin ligase activity as well as BARD1 RING domain binding. These attributes were subsequently used for training a predictive model of homology-directed DNA repair levels for these BRCA1 variants relative to wild type, which is critical for tumor suppression. Here, relative structural changes characterizing BRCA1 variants were quantified by using an efficient and cost-free computational mutagenesis technique, and we show that these features lead to improvements in model performance. This work underscores the potential for bench researchers to gain valuable insights from computational tools, prior to implementing costly and time-consuming experiments.

Phage-DMS: A Comprehensive Method for Fine Mapping of Antibody Epitopes

Understanding the antibody response is critical to developing vaccine and antibody-based therapies and has inspired the recent development of new methods to isolate antibodies. Methods to define the antibody-antigen interactions that determine specificity or allow escape have not kept pace. We developed Phage-DMS, a method that combines two powerful approaches-immunoprecipitation of phage peptide libraries and deep mutational scanning (DMS)-to enable high-throughput fine mapping of antibody epitopes. As an example, we designed sequences encoding all possible amino acid variants of HIV Envelope to create phage libraries. Using Phage-DMS, we identified sites of escape predicted using other approaches for four well-characterized HIV monoclonal antibodies with known linear epitopes. In some cases, the results of Phage-DMS refined the epitope beyond what was determined in previous studies. This method has the potential to rapidly and comprehensively screen many antibodies in a single experiment to define sites essential for binding interactions.

Mutational resilience of antiviral restriction favors primate TRIM5α in host-virus evolutionary arms races

Host antiviral proteins engage in evolutionary arms races with viruses, in which both sides rapidly evolve at interaction interfaces to gain or evade immune defense. For example, primate TRIM5α uses its rapidly evolving 'v1' loop to bind retroviral capsids, and single mutations in this loop can dramatically improve retroviral restriction. However, it is unknown whether such gains of viral restriction are rare, or if they incur loss of pre-existing function against other viruses. Using deep mutational scanning, we comprehensively measured how single mutations in the TRIM5α v1 loop affect restriction of divergent retroviruses. Unexpectedly, we found that the majority of mutations increase weak antiviral function. Moreover, most random mutations do not disrupt potent viral restriction, even when it is newly acquired via a single adaptive substitution. Our results indicate that TRIM5α's adaptive landscape is remarkably broad and mutationally resilient, maximizing its chances of success in evolutionary arms races with retroviruses.

Deep mutagenesis in the study of COVID-19: a technical overview for the proteomics community

INTRODUCTION

The spike (S) of SARS coronavirus 2 (SARS-CoV-2) engages angiotensin-converting enzyme 2 (ACE2) on a host cell to trigger viral-cell membrane fusion and infection. The extracellular region of ACE2 can be administered as a soluble decoy to compete for binding sites on the receptor-binding domain (RBD) of S, but it has only moderate affinity and efficacy. The RBD, which is targeted by neutralizing antibodies, may also change and adapt through mutation as SARS-CoV-2 becomes endemic, posing challenges for therapeutic and vaccine development.

AREAS COVERED

Deep mutagenesis is a Big Data approach to characterizing sequence variants. A deep mutational scan of ACE2 expressed on human cells identified mutations that increase S affinity and guided the engineering of a potent and broad soluble receptor decoy. A deep mutational scan of the RBD displayed on the surface of yeast has revealed residues tolerant of mutational changes that may act as a source for drug resistance and antigenic drift.

EXPERT OPINION

Deep mutagenesis requires a selection of diverse sequence variants; an in vitro evolution experiment that is tracked with next-generation sequencing. The choice of expression system, diversity of the variant library and selection strategy have important consequences for data quality and interpretation.

ZNF423 patient variants, truncations, and in-frame deletions in mice define an allele-dependent range of midline brain abnormalities

Interpreting rare variants remains a challenge in personal genomics, especially for disorders with several causal genes and for genes that cause multiple disorders. ZNF423 encodes a transcriptional regulatory protein that intersects several developmental pathways. ZNF423 has been implicated in rare neurodevelopmental disorders, consistent with midline brain defects in Zfp423-mutant mice, but pathogenic potential of most patient variants remains uncertain. We engineered ~50 patient-derived and small deletion variants into the highly-conserved mouse ortholog and examined neuroanatomical measures for 791 littermate pairs. Three substitutions previously asserted pathogenic appeared benign, while a fourth was effectively null. Heterozygous premature termination codon (PTC) variants showed mild haploabnormality, consistent with loss-of-function intolerance inferred from human population data. In-frame deletions of specific zinc fingers showed mild to moderate abnormalities, as did low-expression variants. These results affirm the need for functional validation of rare variants in biological context and demonstrate cost-effective modeling of neuroanatomical abnormalities in mice.

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.

Reengineering biocatalysts: Computational redesign of chondroitinase ABC improves efficacy and stability

Maintaining biocatalyst stability and activity is a critical challenge. Chondroitinase ABC (ChABC) has shown promise in central nervous system (CNS) regeneration, yet its therapeutic utility is severely limited by instability. We computationally reengineered ChABC by introducing 37, 55, and 92 amino acid changes using consensus design and forcefield-based optimization. All mutants were more stable than wild-type ChABC with increased aggregation temperatures between 4° and 8°C. Only ChABC with 37 mutations (ChABC-37) was more active and had a 6.5 times greater half-life than wild-type ChABC, increasing to 106 hours (4.4 days) from only 16.8 hours. ChABC-37, expressed as a fusion protein with Src homology 3 (ChABC-37-SH3), was active for 7 days when released from a hydrogel modified with SH3-binding peptides. This study demonstrates the broad opportunity to improve biocatalysts through computational engineering and sets the stage for future testing of this substantially improved protein in the treatment of debilitating CNS injuries.

Large-scale DNA-based phenotypic recording and deep learning enable highly accurate sequence-function mapping

Predicting effects of gene regulatory elements (GREs) is a longstanding challenge in biology. Machine learning may address this, but requires large datasets linking GREs to their quantitative function. However, experimental methods to generate such datasets are either application-specific or technically complex and error-prone. Here, we introduce DNA-based phenotypic recording as a widely applicable, practicable approach to generate large-scale sequence-function datasets. We use a site-specific recombinase to directly record a GRE's effect in DNA, enabling readout of both sequence and quantitative function for extremely large GRE-sets via next-generation sequencing. We record translation kinetics of over 300,000 bacterial ribosome binding sites (RBSs) in >2.7 million sequence-function pairs in a single experiment. Further, we introduce a deep learning approach employing ensembling and uncertainty modelling that predicts RBS function with high accuracy, outperforming state-of-the-art methods. DNA-based phenotypic recording combined with deep learning represents a major advance in our ability to predict function from genetic sequence.

Structure-guided engineering of the affinity and specificity of CARs against Tn-glycopeptides

The potency of adoptive T cell therapies targeting the cell surface antigen CD19 has been demonstrated in hematopoietic cancers. It has been difficult to identify appropriate targets in nonhematopoietic tumors, but one class of antigens that have shown promise is aberrant O-glycoprotein epitopes. It has long been known that dysregulated synthesis of O-linked (threonine or serine) sugars occurs in many cancers, and that this can lead to the expression of cell surface proteins containing O-glycans comprised of a single N-acetylgalactosamine (GalNAc, known as Tn antigen) rather than the normally extended carbohydrate. Previously, we used the scFv fragment of antibody 237 as a chimeric antigen receptor (CAR) to mediate recognition of mouse tumor cells that bear its cognate Tn-glycopeptide epitope in podoplanin, also called OTS8. Guided by the structure of the 237 Fab:Tn-OTS8-glycopeptide complex, here we conducted a deep mutational scan showing that residues flanking the Tn-glycan contributed significant binding energy to the interaction. Design of 237-scFv libraries in the yeast display system allowed us to isolate scFv variants with higher affinity for Tn-OTS8. Selection with a noncognate human antigen, Tn-MUC1, yielded scFv variants that were broadly reactive with multiple Tn-glycoproteins. When configured as CARs, engineered T cells expressing these scFv variants showed improved activity against mouse and human cancer cell lines defective in O-linked glycosylation. This strategy provides CARs with Tn-peptide specificities, all based on a single scFv scaffold, that allows the same CAR to be tested for toxicity in mice and efficacy against mouse and human tumors.

Peptide Folding and Binding Probed by Systematic Non-canonical Mutagenesis

Many proteins and peptides fold upon binding another protein. Mutagenesis has proved an essential tool in the study of these multi-step molecular recognition processes. By comparing the biophysical behavior of carefully selected mutants, the concert of interactions and conformational changes that occur during folding and binding can be separated and assessed. Recently, this mutagenesis approach has been radically expanded by deep mutational scanning methods, which allow for many thousands of mutations to be examined in parallel. Furthermore, these high-throughput mutagenesis methods have been expanded to include mutations to non-canonical amino acids, returning peptide structure-activity relationships with unprecedented depth and detail. These developments are timely, as the insights they provide can guide the optimization of de novo cyclic peptides, a promising new modality for chemical probes and therapeutic agents.

Improved mutant function prediction via PACT: Protein Analysis and Classifier Toolkit

MOTIVATION

Deep mutational scanning experiments have enabled the measurement of the sequence-function relationship for thousands of mutations in a single experiment. The Protein Analysis and Classifier Toolkit (PACT) is a Python software package that marries the fitness metric of a given mutation within these experiments to sequence and structural features enabling downstream analyses. PACT enables the easy development of user sharable protocols for custom deep mutational scanning experiments as all code is modular and reusable between protocols. Protocols for mutational libraries with single or multiple mutations are included. To exemplify its utility, PACT assessed two deep mutational scanning datasets that measured the tradeoff of enzyme activity and enzyme stability.

RESULTS

PACT efficiently evaluated classifiers that predict protein mutant function tested on deep mutational scanning screens. We found that the classifiers with the lowest false positive and highest true positive rate assesses sequence homology, contact number and if mutation involves proline.

AVAILABILITY AND IMPLEMENTATION

PACT and the processed datasets are distributed freely under the terms of the GPL-3 license. The source code is available at GitHub (https://github.com/JKlesmith/PACT).

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Comprehensive exploration of the translocation, stability and substrate recognition requirements in VIM-2 lactamase

Metallo-β-lactamases (MBLs) degrade a broad spectrum of β-lactam antibiotics, and are a major disseminating source for multidrug resistant bacteria. Despite many biochemical studies in diverse MBLs, molecular understanding of the roles of residues in the enzyme's stability and function, and especially substrate specificity, is lacking. Here, we employ deep mutational scanning (DMS) to generate comprehensive single amino acid variant data on a major clinical MBL, VIM-2, by measuring the effect of thousands of VIM-2 mutants on the degradation of three representative classes of β-lactams (ampicillin, cefotaxime, and meropenem) and at two different temperatures (25°C and 37°C). We revealed residues responsible for expression and translocation, and mutations that increase resistance and/or alter substrate specificity. The distribution of specificity-altering mutations unveiled distinct molecular recognition of the three substrates. Moreover, these function-altering mutations are frequently observed among naturally occurring variants, suggesting that the enzymes have continuously evolved to become more potent resistance genes.

Predicting the viability of beta-lactamase: How folding and binding free energies correlate with beta-lactamase fitness

One of the long-standing holy grails of molecular evolution has been the ability to predict an organism's fitness directly from its genotype. With such predictive abilities in hand, researchers would be able to more accurately forecast how organisms will evolve and how proteins with novel functions could be engineered, leading to revolutionary advances in medicine and biotechnology. In this work, we assemble the largest reported set of experimental TEM-1 β-lactamase folding free energies and use this data in conjunction with previously acquired fitness data and computational free energy predictions to determine how much of the fitness of β-lactamase can be directly predicted by thermodynamic folding and binding free energies. We focus upon β-lactamase because of its long history as a model enzyme and its central role in antibiotic resistance. Based upon a set of 21 β-lactamase single and double mutants expressly designed to influence protein folding, we first demonstrate that modeling software designed to compute folding free energies such as FoldX and PyRosetta can meaningfully, although not perfectly, predict the experimental folding free energies of single mutants. Interestingly, while these techniques also yield sensible double mutant free energies, we show that they do so for the wrong physical reasons. We then go on to assess how well both experimental and computational folding free energies explain single mutant fitness. We find that folding free energies account for, at most, 24% of the variance in β-lactamase fitness values according to linear models and, somewhat surprisingly, complementing folding free energies with computationally-predicted binding free energies of residues near the active site only increases the folding-only figure by a few percent. This strongly suggests that the majority of β-lactamase's fitness is controlled by factors other than free energies. Overall, our results shed a bright light on to what extent the community is justified in using thermodynamic measures to infer protein fitness as well as how applicable modern computational techniques for predicting free energies will be to the large data sets of multiply-mutated proteins forthcoming.

funtrp: identifying protein positions for variation driven functional tuning

Evaluating the impact of non-synonymous genetic variants is essential for uncovering disease associations and mechanisms of evolution. An in-depth understanding of sequence changes is also fundamental for synthetic protein design and stability assessments. However, the variant effect predictor performance gain observed in recent years has not kept up with the increased complexity of new methods. One likely reason for this might be that most approaches use similar sets of gene and protein features for modeling variant effects, often emphasizing sequence conservation. While high levels of conservation highlight residues essential for protein activity, much of the variation observable in vivo is arguably weaker in its impact, thus requiring evaluation at a higher level of resolution. Here, we describe functionNeutral/Toggle/Rheostatpredictor (funtrp), a novel computational method that categorizes protein positions based on the position-specific expected range of mutational impacts: Neutral (weak/no effects), Rheostat (function-tuning positions), or Toggle (on/off switches). We show that position types do not correlate strongly with familiar protein features such as conservation or protein disorder. We also find that position type distribution varies across different protein functions. Finally, we demonstrate that position types can improve performance of existing variant effect predictors and suggest a way forward for the development of new ones.

Minimum epistasis interpolation for sequence-function relationships

Massively parallel phenotyping assays have provided unprecedented insight into how multiple mutations combine to determine biological function. While such assays can measure phenotypes for thousands to millions of genotypes in a single experiment, in practice these measurements are not exhaustive, so that there is a need for techniques to impute values for genotypes whose phenotypes have not been directly assayed. Here, we present an imputation method based on inferring the least epistatic possible sequence-function relationship compatible with the data. In particular, we infer the reconstruction where mutational effects change as little as possible across adjacent genetic backgrounds. The resulting models can capture complex higher-order genetic interactions near the data, but approach additivity where data is sparse or absent. We apply the method to high-throughput transcription factor binding assays and use it to explore a fitness landscape for protein G.

Emerging strategies to bridge the gap between pharmacogenomic research and its clinical implementation

The genomic inter-individual heterogeneity remains a significant challenge for both clinical decision-making and the design of clinical trials. Although next-generation sequencing (NGS) is increasingly implemented in drug development and clinical trials, translation of the obtained genomic information into actionable clinical advice lags behind. Major reasons are the paucity of sufficiently powered trials that can quantify the added value of pharmacogenetic testing, and the considerable pharmacogenetic complexity with millions of rare variants with unclear functional consequences. The resulting uncertainty is reflected in inconsistencies of pharmacogenomic drug labels in Europe and the United States. In this review, we discuss how the knowledge gap for bridging pharmacogenomics into the clinics can be reduced. First, emerging methods that allow the high-throughput experimental characterization of pharmacogenomic variants combined with novel computational tools hold promise to improve the accuracy of drug response predictions. Second, tapping of large biobanks of therapeutic drug monitoring data allows to conduct high-powered retrospective studies that can validate the clinical importance of genetic variants, which are currently incompletely characterized. Combined, we are confident that these methods will improve the accuracy of drug response predictions and will narrow the gap between variant identification and its utilization for clinical decision-support.

Comprehensive fitness maps of Hsp90 show widespread environmental dependence

Gene-environment interactions have long been theorized to influence molecular evolution. However, the environmental dependence of most mutations remains unknown. Using deep mutational scanning, we engineered yeast with all 44,604 single codon changes encoding 14,160 amino acid variants in Hsp90 and quantified growth effects under standard conditions and under five stress conditions. To our knowledge, these are the largest determined comprehensive fitness maps of point mutants. The growth of many variants differed between conditions, indicating that environment can have a large impact on Hsp90 evolution. Multiple variants provided growth advantages under individual conditions; however, these variants tended to exhibit growth defects in other environments. The diversity of Hsp90 sequences observed in extant eukaryotes preferentially contains variants that supported robust growth under all tested conditions. Rather than favoring substitutions in individual conditions, the long-term selective pressure on Hsp90 may have been that of fluctuating environments, leading to robustness under a variety of conditions.

A proactive genotype-to-patient-phenotype map for cystathionine beta-synthase

BACKGROUND

For the majority of rare clinical missense variants, pathogenicity status cannot currently be classified. Classical homocystinuria, characterized by elevated homocysteine in plasma and urine, is caused by variants in the cystathionine beta-synthase (CBS) gene, most of which are rare. With early detection, existing therapies are highly effective.

METHODS

Damaging CBS variants can be detected based on their failure to restore growth in yeast cells lacking the yeast ortholog CYS4. This assay has only been applied reactively, after first observing a variant in patients. Using saturation codon-mutagenesis, en masse growth selection, and sequencing, we generated a comprehensive, proactive map of CBS missense variant function.

RESULTS

Our CBS variant effect map far exceeds the performance of computational predictors of disease variants. Map scores correlated strongly with both disease severity (Spearman's ϱ = 0.9) and human clinical response to vitamin B6 (ϱ = 0.93).

CONCLUSIONS

We demonstrate that highly multiplexed cell-based assays can yield proactive maps of variant function and patient response to therapy, even for rare variants not previously seen in the clinic.

Development and Application of Computational Methods in Phage Display Technology

BACKGROUND

Phage display is a powerful and versatile technology for the identification of peptide ligands binding to multiple targets, which has been successfully employed in various fields, such as diagnostics and therapeutics, drug-delivery and material science. The integration of next generation sequencing technology with phage display makes this methodology more productive. With the widespread use of this technique and the fast accumulation of phage display data, databases for these data and computational methods have become an indispensable part in this community. This review aims to summarize and discuss recent progress in the development and application of computational methods in the field of phage display.

METHODS

We undertook a comprehensive search of bioinformatics resources and computational methods for phage display data via Google Scholar and PubMed. The methods and tools were further divided into different categories according to their uses.

RESULTS

We described seven special or relevant databases for phage display data, which provided an evidence-based source for phage display researchers to clean their biopanning results. These databases can identify and report possible target-unrelated peptides (TUPs), thereby excluding false-positive data from peptides obtained from phage display screening experiments. More than 20 computational methods for analyzing biopanning data were also reviewed. These methods were classified into computational methods for reporting TUPs, for predicting epitopes and for analyzing next generation phage display data.

CONCLUSION

The current bioinformatics archives, methods and tools reviewed here have benefitted the biopanning community. To develop better or new computational tools, some promising directions are also discussed.

Deep2Full: Evaluating strategies for selecting the minimal mutational experiments for optimal computational predictions of deep mutational scan outcomes

Performing a complete deep mutational scan with all single point mutations may not be practical, and may not even be required, especially if predictive computational models can be developed. Computational models are however naive to cellular response in the myriads of assay-conditions. In a realistic paradigm of assay context-aware predictive hybrid models that combine minimal experimental data from deep mutational scans with structure, sequence information and computational models, we define and evaluate different strategies for choosing this minimal set. We evaluated the trivial strategy of a systematic reduction in the number of mutational studies from 85% to 15%, along with several others about the choice of the types of mutations such as random versus site-directed with the same 15% data completeness. Interestingly, the predictive capabilities by training on a random set of mutations and using a systematic substitution of all amino acids to alanine, asparagine and histidine (ANH) were comparable. Another strategy we explored, augmenting the training data with measurements of the same mutants at multiple assay conditions, did not improve the prediction quality. For the six proteins we analyzed, the bin-wise error in prediction is optimal when 50-100 mutations per bin are used in training the computational model, suggesting that good prediction quality may be achieved with a library of 500-1000 mutations.

Algorithm-based coevolution network identification reveals key functional residues of the α/β hydrolase subfamilies

Covariant residues identified by computational algorithms have provided new insights into enzyme evolutionary routes. However, the reliability and accuracy of routine statistical coupling analysis (SCA) are unable to satisfy the needs of protein engineering because SCA depends only on sequence information. Here, we set up a new SCA algorithm, SCA.SIM, by integrating structure information and MD simulation data. The more reliable covariant residues with high-quality scores are obtained from sequence alignment weighted by residual movement for eight related subfamilies, belonging to α/β hydrolase family, with Candida antarctica lipase B (CALB). The 38 predicted covariant residues are tested for function by high-throughput quantitative evaluation in combination with activity and thermostability assays of a mutant library and deep sequencing. Based on the landscapes of both activity and thermostability, most mutants play key roles in catalysis, and some mutants gain 2.4- to 6-fold increase in half-life at 50°C and 9- to 12-fold improvement in catalytic efficiency. The activity of double mutants for A225F/T103A is higher than those of A225F and T103A which means that SCA.SIM method might be useful for identifying the allosteric coupling. The SCA.SIM algorithm can be used for protein coevolution and enzyme engineering research.

Design of a basigin-mimicking inhibitor targeting the malaria invasion protein RH5

Many human pathogens use host cell-surface receptors to attach and invade cells. Often, the host-pathogen interaction affinity is low, presenting opportunities to block invasion using a soluble, high-affinity mimic of the host protein. The Plasmodium falciparum reticulocyte-binding protein homolog 5 (RH5) provides an exciting candidate for mimicry: it is highly conserved and its moderate affinity binding to the human receptor basigin (KD  ≥1 μM) is an essential step in erythrocyte invasion by this malaria parasite. We used deep mutational scanning of a soluble fragment of human basigin to systematically characterize point mutations that enhance basigin affinity for RH5 and then used Rosetta to design a variant within the sequence space of affinity-enhancing mutations. The resulting seven-mutation design exhibited 1900-fold higher affinity (KD approximately 1 nM) for RH5 with a very slow binding off rate (0.23 h-1 ) and reduced the effective Plasmodium growth-inhibitory concentration by at least 10-fold compared to human basigin. The design provides a favorable starting point for engineering on-rate improvements that are likely to be essential to reach therapeutically effective growth inhibition.

Analyzing the Sequences of Intrinsically Disordered Regions with CIDER and localCIDER

Intrinsically disordered proteins and protein regions are ubiquitous across eukaryotic proteomes where they play a range of functional roles. Unlike folded proteins, IDRs lack a well-defined native state but exist in heterogeneous ensembles of conformations. In the absence of a defined native state, structure-guided mutations to test specific mechanistic hypotheses are generally not possible. Despite this, the use of mutations to alter sequence properties has become a relatively common approach for teasing out the relationship between sequence, ensemble, and function. A key step in designing informative mutants is the ability to identify specific sequence features that may reveal an interpretable response if perturbed. Here, we provide guidance on using the CIDER and localCIDER tools for amino acid sequence analysis, with a focus on building intuition with respect to the most commonly described features.

A Phage-Assisted Continuous Selection Approach for Deep Mutational Scanning of Protein-Protein Interactions

Protein-protein interactions (PPIs) are critical for organizing molecules in a cell and mediating signaling pathways. Dysregulation of PPIs is often a key driver of disease. To better understand the biophysical basis of such disease processes-and to potentially target them-it is critical to understand the molecular determinants of PPIs. Deep mutational scanning (DMS) facilitates the acquisition of large amounts of biochemical data by coupling selection with high throughput sequencing (HTS). The challenging and labor-intensive design and optimization of a relevant selection platform for DMS, however, limits the use of powerful directed evolution and selection approaches. To address this limitation, we designed a versatile new phage-assisted continuous selection (PACS) system using our previously reported proximity-dependent split RNA polymerase (RNAP) biosensors, with the aim of greatly simplifying and streamlining the design of a new selection platform for PPIs. After characterization and validation using the model KRAS/RAF PPI, we generated a library of RAF variants and subjected them to PACS and DMS. Our HTS data revealed positions along the binding interface that are both tolerant and intolerant to mutations, as well as which substitutions are tolerated at each position. Critically, the "functional scores" obtained from enrichment data through continuous selection for individual variants correlated with K_D values measured in vitro, indicating that biochemical data can be extrapolated from sequencing using our new system. Due to the plug and play nature of RNAP biosensors, this method can likely be extended to a variety of other PPIs. More broadly, this, and other methods under development support the continued development of evolutionary and high-throughput approaches to address biochemical problems, moving toward a more comprehensive understanding of sequence-function relationships in proteins.

Recommendations for the collection and use of multiplexed functional data for clinical variant interpretation

Variants of uncertain significance represent a massive challenge to medical genetics. Multiplexed functional assays, in which the functional effects of thousands of genomic variants are assessed simultaneously, are increasingly generating data that can be used as additional evidence for or against variant pathogenicity. Such assays have the potential to resolve variants of uncertain significance, thereby increasing the clinical utility of genomic testing. Existing standards from the American College of Medical Genetics and Genomics (ACMG)/Association for Molecular Pathology (AMP) and new guidelines from the Clinical Genome Resource (ClinGen) establish the role of functional data in variant interpretation, but do not address the specific challenges or advantages of using functional data derived from multiplexed assays. Here, we build on these existing guidelines to provide recommendations to experimentalists for the production and reporting of multiplexed functional data and to clinicians for the evaluation and use of such data. By following these recommendations, experimentalists can produce transparent, complete, and well-validated datasets that are primed for clinical uptake. Our recommendations to clinicians and diagnostic labs on how to evaluate the quality of multiplexed functional datasets, and how different datasets could be incorporated into the ACMG/AMP variant-interpretation framework, will hopefully clarify whether and how such data should be used. The recommendations that we provide are designed to enhance the quality and utility of multiplexed functional data, and to promote their judicious use.