Biology subjects

Keating, A. E.

Publications and source records attributed to Keating, A. E..

15 recordsLinked to original sources

bioRxiv2026

Seed-Guided De Novo Design Expands the Structural Diversity of Antitoxin Protein Binders

De novo design of protein binders targeting extended, multi-site interaction surfaces remains difficult for current generative methods, which often produce limited structural diversity and predominantly helical topologies. Here, we advance diffusion-based binder design by guiding protein backbone generation through "seeds," which are PDB-derived fragments selected for geometric complementarity to the target surface. To test this approach, we computationally generated seed-guided binders of the bacterial toxin RelE. RelB, the native antitoxin of RelE, engages two distinct interfaces with high surface complementarity, making it an appropriate test case. Seed-guided RFdiffusion produced backbones with substantially higher structural diversity and more target contacts than RFdiffusion alone. Experimental screening of 1,402 designs in a high-throughput bacterial survival assay identified multiple functional binders, including variants with nanomolar to low-micromolar affinity and one design with RelE neutralization comparable to RelBpep. Computational structure prediction and mutational analyses support that the designed interfaces rely on seed-derived contacts and adopt binding modes distinct from RelB. Molecular dynamics simulations and hydrogen-deuterium exchange experiments further suggest that one high-affinity design undergoes a conformational change upon binding. Notably, successful designs exhibited reduced cross-reactivity to RelE orthologs compared with RelBpep, suggesting that the extensive interfaces generated through seed-guided design can enable enhanced selectivity. These results establish motif scaffolding of surface-complementing seeds as an effective strategy for overcoming current limitations of de novo generative models, enabling the design of proteins that can engage challenging interface sites.

bioengineering↗

bioRxiv2026

Multiomic screening platform uncovers the impact of histone mutations on chromatin and cell fate

Somatic missense mutations in histone genes, often referred to as oncohistones, have been identified in diverse types of human cancers. The functional and mechanistic impact of most oncohistones remains unknown. To address this gap, we developed CHANCLA, a modular platform for high-throughput functional screening of oncohistones using multiomic phenotypic readouts. We used CHANCLA to systematically measure the impact of 303 human oncohistones on cellular proliferation, differentiation, histone-specific post-translational modifications, and chromatin accessibility. Integrative multiomic analyses revealed discrete oncohistone molecular classes that promote proliferation, block lineage-specific differentiation, and physically remodel the chromatin landscape by altering specific histone modifications and reducing nucleosome stability. Structural mapping and computational modeling studies uncovered that functionally convergent mutations are clustered at key nucleosome interfaces, particularly H2B-H4, and that chromatin accessibility-promoting mutations are linked to mono-nucleosome destabilization. Leveraging this multiomic resource, we discovered that the H3.3-Q5H mutant histone is a bona fide human oncohistone that accelerates lung adenocarcinoma growth in vivo. Mechanistically, we found that H3.3-Q5H expression leads to suppression of promoter-associated H3K4me3 and expansion of repressive H3K27me3 domains, resulting in increased KRAS signaling and gene expression programs associated with epithelial-to-mesenchymal transition. Together, this work provides a multiomic functional atlas of cancer-associated histone mutations, identifies structural and mechanistic principles governing chromatin reprogramming by oncohistones, and establishes CHANCLA as a modular platform for systematic discovery of mechanisms and vulnerabilities associated with these genetic lesions.

cancer biology↗

bioRxiv2026

Beyond native sequence recovery: Improved modeling of thesequence-energy landscape of protein structures

Computational protein design using machine learning models has advanced rapidly since the introduction of AlphaFold2. There is now a suite of tools that enable in silico design of proteins with desired structures and properties. Most design workflows require fitting a designed backbone with a sequence that stabilizes it, and many machine learning sequence design models have been proposed. These models are trained to recover the native sequence paired with a known structure, a task known as native sequence recovery (NSR). Here, we demonstrate the limitations of optimizing a sequence design model only for NSR. We show that NSR is often misaligned with more important metrics of model performance: the compatibility of the generated sequence with the desired fold and the ability of the model to predict the energetic effects of mutations. We introduce PottsMPNN, which is trained to generate a Potts energy function consisting of single-residue and residue-pair terms from a protein backbone, and we demonstrate that learning a Potts model reduces NSR but improves sequence generation and energy prediction. To further show that NSR is not the optimal metric, we trained PottsMPNN with noised backbone structures and multiple sequence alignments. In tests on held-out data, NSR decreased, but the quality of the designed sequences and energy predictions improved. By demonstrating the limitations of optimizing for NSR and the effectiveness of alternative strategies for avoiding over-optimizing for NSR, our work provides a new direction for the sequence design field.

biophysics↗

bioRxiv2026

A proteome-wide biochemical screen defines binding determinants of the core autophagy protein LC3B

Human MAP1LC3B (LC3B) binds proteins involved in autophagy and other cellular processes using a degenerate four-residue short linear motif known as the LC3-interacting region (LIR). Biochemical and structural studies have identified LIRs in many LC3B interaction partners, but the sequence features that contribute to binding have not been systematically explored. To discover peptides that interact with LC3B and deeply profile the key binding determinants, we screened a library of ~500,000 36-residue peptides derived from the human proteome using bacterial cell-surface display. Analysis of the screening data, coupled with structural studies and site-directed mutagenesis, revealed exceptions to the canonical LIR motif ([FWY]0-X1-X2-[LVI]3) and a strong preference for negatively charged residues adjacent to the LIR, which we visualized in a newly determined structure of LC3B bound to a peptide isolated in our screen. Guided by our screening data, we designed synthetic LIR-containing peptides that bind LC3B with affinities comparable to the tightest measured natural binder. Finally, we determined that mutations in LC3B commonly thought to abrogate binding of LIR-containing peptides instead alter LC3B binding specificity, leading to enhanced binding of some LIR-containing sequences. Taken together, our results refine the LIR motif definition, expand the network of candidate LC3B interaction partners, and highlight how mutations at the LIR docking site can modulate affinity and specificity.

biochemistry↗

bioRxiv2025

How AlphaFold and related models predict protein-peptide complex structures

Protein-peptide interactions mediate many biological processes, and access to accurate structural models, through experimental determination or reliable computational prediction, is essential for understanding protein function and designing novel protein-protein interactions. AlphaFold2-Multimer (AF2-Multimer), AlphaFold3 (AF3), and related models such as Boltz-1 and Chai-1 are state-of-the-art protein structure predictors that successfully predict protein-peptide complex structures. Using a dataset of experimentally resolved protein-peptide structures, we analyzed the performance of these four structure prediction models to understand how they work. We found evidence of bias for previously seen structures, suggesting that models may struggle to generalize to novel target proteins or binding sites. We probed how models use the protein and peptide multiple sequence alignments (MSAs), which are often shallow or of poor quality for peptide sequences. We found weak evidence that models use coevolutionary information from paired MSAs and found that both the target and peptide unpaired MSAs contribute to performance. Our work highlights the promise of deep learning for peptide docking and the importance of diverse representation of interface geometries in the training data for optimal prediction performance.

bioinformatics↗

bioRxiv2024

STAT5B leukemic mutations, altering SH2 tyrosine 665, have opposing impacts on immune gene programs

STAT5B is a vital transcription factor for lymphocytes. Here, function of two STAT5B mutations from human T cell leukemias: one substituting tyrosine 665 with phenylalanine (STAT5BY665F), the other with histidine (STAT5BY665H) was interrogated. In silico modeling predicted divergent energetic effects on homodimerization with a range of pathogenicity. In primary T cells in vitro STAT5BY665F showed gain-of-function while STAT5BY665H demonstrated loss-of-function. Introducing the mutation into the mouse genome illustrated that the gain-of-function Stat5bY665F mutation resulted in accumulation of CD8+ effector and memory and CD4+ regulatory T-cells, altering CD8+/CD4+ ratios. In contrast, STAT5BY665H knock-in mice showed diminished CD8+ effector and memory and CD4+ regulatory T cells. In contrast to wild-type STAT5, the STAT5BY665F variant displayed greater STAT5 phosphorylation, DNA binding and transcriptional activity following cytokine activation while the STAT5BY665H variant resembled a null. The work exemplifies how joining in silico and in vivo studies of single nucleotides deepens our understanding of disease-associated variants, revealing structural determinants of altered function, defining mechanistic roles, and, specifically here, identifying a gain-of function variant that does not directly induce hematopoietic malignancy.

genetics↗

bioRxiv2024

PairK: Pairwise k-mer alignment for quantifying protein motif conservation in disordered regions

Protein-protein interactions are often mediated by a modular peptide recognition domain binding to a short linear motif (SLiM) in the disordered region of another protein. The ability to predict domain-SLiM interactions would allow researchers to map protein interaction networks, predict the effects of perturbations to those networks, and develop biologically meaningful hypotheses. Unfortunately, sequence database searches for SLiMs generally yield mostly biologically irrelevant motif matches or false positives. To improve the prediction of novel SLiM interactions, researchers employ filters to discriminate between biologically relevant and improbable motif matches. One promising criterion for identifying biologically relevant SLiMs is the sequence conservation of the motif, exploiting the fact that functional motifs are more likely to be conserved than spurious motif matches. However, the difficulty of aligning disordered regions has significantly hampered the utility of this approach. We present PairK (pairwise k-mer alignment), an MSA-free method to quantify motif conservation in disordered regions. PairK outperforms both standard MSA-based conservation scores and a modern LLM-based conservation score predictor on the task of identifying biologically important motif instances. PairK can quantify conservation over wider phylogenetic distances than MSAs, indicating that SLiMs may be more conserved than is implied by MSA-based metrics. PairK is available as open-source code at https://github.com/jacksonh1/pairk.

bioinformatics↗

bioRxiv2024

Elaboration of the Homer1 Recognition Landscape Reveals Incomplete Divergence of Paralogous EVH1 Domains

Short sequences that mediate interactions with modular binding domains are ubiquitous throughout eukaryotic proteomes. Networks of Short Linear Motifs (SLiMs) and their corresponding binding domains orchestrate many cellular processes, and the low mutational barrier to evolving novel interactions provides a way for biological systems to rapidly sample selectable phenotypes. Mapping SLiM binding specificity and the rules that govern SLiM evolution is fundamental to uncovering the pathways regulated by these networks and developing the tools to manipulate them. We used high-throughput screening of the human proteome to identify sequences that bind to the Enabled/VASP homology 1 (EVH1) domain of the postsynaptic density scaffolding protein Homer1. In doing so, we expanded current understanding of the determinants of Homer EVH1 binding preferences and defined a new motif that can facilitate the discovery of additional Homer-mediated interactions. Interestingly, the Homer1 EVH1 domain preferentially binds to sequences containing an N-terminally overlapping motif that is bound by the paralogous family of Ena/VASP actin polymerases, and many of these sequences can bind to EVH1 domains from both protein families. We provide evidence from orthologous EVH1 domains in pre-metazoan organisms that the overlap in human Ena/VASP and Homer binding preferences corresponds to an incomplete divergence from a common Ena/VASP ancestor. Given this overlap in binding profiles, promiscuous sequences that can be recognized by both families either achieve specificity through extrinsic regulatory strategies or may provide functional benefits via multi-specificity. This may explain why these paralogs incompletely diverged despite the accessibility of further diverged isoforms. Impact StatementShort linear motifs bind to structurally conserved domains to mediate many protein interactions. Current definitions of motifs, using regular expressions, are over simplified and do not capture what is required for binding. Unbiased proteome interaction screens can refine these definitions. For EVH1 domains considered in this work, screening elucidated Homer1 binding requirements and revealed a residual overlap of specificity of two anciently diverged families, highlighting both opportunities and challenges of multi-specificity in domain-motif interactions.

biochemistry↗

bioRxiv2023

High-throughput computational discovery of inhibitory protein fragments with AlphaFold

Peptides can bind to specific sites on larger proteins and thereby function as inhibitors and regulatory elements. Peptide fragments of larger proteins are particularly attractive for achieving these functions due to their inherent potential to form native-like binding interactions. Recently developed experimental approaches allow for high-throughput measurement of protein fragment inhibitory activity in living cells. However, it has thus far not been possible to predict de novo which of the many possible protein fragments bind to protein targets, let alone act as inhibitors. We have developed a computational method, FragFold, that employs AlphaFold to predict protein fragment binding to full-length proteins in a high-throughput manner. Applying FragFold to thousands of fragments tiling across diverse proteins revealed peaks of predicted binding along each protein sequence. Comparisons with experimental measurements establish that our approach is a sensitive predictor of fragment function: Evaluating inhibitory fragments from known protein-protein interaction interfaces, we find 87% are predicted by FragFold to bind in a native-like mode. Across full protein sequences, 68% of FragFold-predicted binding peaks match experimentally measured inhibitory peaks. Deep mutational scanning experiments support the predicted binding modes and uncover superior inhibitory peptides in high throughput. Further, FragFold is able to predict previously unknown protein binding modes, explaining prior genetic and biochemical data. The success rate of FragFold demonstrates that this computational approach should be broadly applicable for discovering inhibitory protein fragments across proteomes. Significance StatementPeptides can regulate protein interactions by binding to specific interfaces, and fragments of larger proteins have high potential to function in this manner. Recently developed experimental methods allow massively parallel measurement of protein fragment-based inhibition in vivo. However, we have lacked comparable computational methods to predict which protein fragments act as inhibitors and how they bind. Here we report a new approach, FragFold, which leverages high-throughput AlphaFold predictions of protein - fragment binding to tackle these problems at scale. FragFold is successful at predicting inhibitory protein fragments and their binding modes across diverse protein structures and functions. This new approach stands to enable proteome-wide discovery of inhibitory protein fragments and aid the interpretation of high-throughput experimental measurements of inhibitory activity. ClassificationBiological Sciences / Biophysics and Computational Biology

biophysics↗

bioRxiv2023

Marginal specificity in protein interactions constrains evolution

The evolution of novel functions in biology relies heavily on gene duplication and divergence, creating large paralogous protein families. Selective pressure to avoid detrimental cross-talk often results in paralogs that exhibit exquisite specificity for their interaction partners. But how robust or sensitive is this specificity to mutation? Here, using deep mutational scanning, we demonstrate that a paralogous family of bacterial signaling proteins exhibits marginal specificity, such that many individual substitutions give rise to substantial cross-talk between normally insulated pathways. Our results indicate that sequence space is locally crowded despite overall sparseness, and we provide evidence that this crowding has constrained the evolution of bacterial signaling proteins. These findings underscore how evolution selects for good enough rather than optimized phenotypes, leading to restrictions on the subsequent evolvability of paralogs. Significance StatementLarge paralogous protein families are found throughout biology, the product of extensive gene duplication. To execute different functions inside cells, paralogs typically acquire different specificities, interacting with only desired, cognate partners and avoiding cross-talk with non-cognate proteins. But how robust is this interaction specificity to mutation? Can individual mutations lead to cross-talk or do paralogs diverge enough such that multiple mutations would be required, providing a mutational buffer against cross-talk? To address these questions, we built mutant libraries that produce all possible single substitutions of a bacterial kinase and then screened for cross-talk to non-cognate proteins. Strikingly, we find that many single substitutions can produce cross-talk, meaning that these pathways typically exhibit only marginal specificity, and demonstrate that this restricts their evolvability.

evolutionary biology↗

bioRxiv2022

Neural Network-Derived Potts Models for Structure-Based Protein Design using Backbone Atomic Coordinates and Tertiary Motifs

AO_SCPLOWBSTRACTC_SCPLOWDesigning novel proteins to perform desired functions, such as binding or catalysis, is a major goal in synthetic biology. A variety of computational approaches can aid in this task. An energy-based framework rooted in the sequence-structure statistics of tertiary motifs (TERMs) can be used for sequence design on pre-defined backbones. Neural network models that use backbone coordinate-derived features provide another way to design new proteins. In this work, we combine the two methods to make neural structure-based models more suitable for protein design. Specifically, we supplement backbone-coordinate features with TERM-derived data, as inputs, and we generate energy functions as outputs. We present two architectures that generate Potts models over the sequence space: TERMinator, which uses both TERM-based and coordinate-based information, and COORDinator, which uses only coordinate-based information. Using these two models, we demonstrate that TERMs can be utilized to improve native sequence recovery performance of neural models. Furthermore, we demonstrate that sequences designed by TERMinator are predicted to fold to their target structures by AlphaFold. Finally, we show that both TERMinator and COORDinator learn notions of energetics, and these methods can be fine-tuned on experimental data to improve predictions. Our results suggest that using TERM-based and coordinate-based features together may be beneficial for protein design and that structure-based neural models that produce Potts energy tables have utility for flexible applications in protein science. CodeCode will be made publically available at https://github.com/alexjli/terminator_public

biophysics↗

bioRxiv2022

High-throughput discovery of TRAF6-interacting peptides identifies determinants of positive and negative design and shows known and candidate human interaction partner motifs are not optimized for affinity

TRAF6 is an adapter protein and E3 ubiquitin ligase involved in signaling downstream of cell receptors essential for development and the immune system. TRAF6 participates in many protein-protein interactions, some of which are mediated by a C-terminal MATH domain that recruits TRAF6 to cell-surface receptors and associated proteins. The TRAF6 MATH domain binds to short peptide segments containing the motif PxExx[FYWHDE], where x is any amino acid. Blocking TRAF6 interactions is associated with favorable effects in various disease models. To better define the TRAF6 MATH domain binding preferences, we generated a bacterial cell-surface peptide display library to sample the TRAF6 motif sequence space. We performed sorting experiments and identified 236 of the best TRAF6-interacting peptides and a set of 1,200 peptides that match the sequence PxE but do not bind TRAF6. Selected binders, tested by single-clone bacterial display titrations and bio-layer interferometry, bound TRAF6 tighter than previously measured native peptides. To elucidate the structural basis for TRAF6 interaction preferences, we built all-atom structural models of the TRAF6 MATH domain in complex with high-affinity binders and motif-matching nonbinders that were identified in the screen. We identified motif features that favor binding to TRAF6 as well as negative design elements distributed across the motif that can disfavor or preclude binding. Searching the human proteome for matches to the library screening-defined binding motif revealed that most known, biologically relevant TRAF6 motif matches occupy a different sequence space from the most enriched hits discovered in combinatorial library screening. Our experimentally determined binding preferences and structural models can support the design of peptide-based interaction inhibitors with higher affinities than endogenous TRAF6 ligands.

biochemistry↗

bioRxiv2021

A distributed residue network permits conformational binding specificity in a conserved family of actin remodelers

Metazoan proteomes contain many paralogous proteins that have evolved distinct functions. The Ena/VASP family of actin regulators consists of three members that share an EVH1 interaction domain with a 100% conserved binding site. A proteome-wide screen revealed ciliary protein PCARE as a high-affinity ligand for ENAH EVH1. Here we report the surprising observation that PCARE is ~100-fold specific for ENAH over paralogs VASP and EVL and can selectively bind and inhibit ENAH-dependent adhesion in cells. Specificity arises from a mechanism whereby PCARE stabilizes a conformation of the ENAH EVH1 domain that is inaccessible to family members VASP and EVL. Structure-based modeling rapidly identified seven residues distributed throughout EVL that are sufficient to differentiate binding by ENAH vs. EVL. By exploiting the ENAH-specific conformation, we rationally designed the tightest and most selective ENAH binder to date. Our work uncovers a conformational mechanism of interaction specificity that distinguishes highly similar paralogs and establishes tools for dissecting specific Ena/VASP functions in processes including cancer cell invasion.

biochemistry↗

bioRxiv2021

A proteome-wide screen uncovers diverse roles for sequence context surrounding proline-rich motifs in Ena/VASP molecular recognition

The human proteome is replete with short linear motifs (SLiMs) of 4-6 residues that are critical for protein-protein interactions, yet the importance of the sequence surrounding such motifs is underexplored. We devised a proteomic screen to systematically examine the influence of SLiM sequence context on protein-protein interactions. Focusing on the EVH1 domain of ENAH, an actin regulator that is upregulated in invasive cancers, we screened 36-residue proteome-derived peptides for binding. We discovered a pocket on the ENAH EVH1 domain that diverged from its orthologs to recognize extended SLiMs, and we found that proteins with two EVH1-binding SLiMs can wrap around a single domain. We also found that the ciliary protein PCARE uses an extended 23-residue region to obtain higher affinity than any known ENAH EVH1-binding motif. Our screen provides a way to uncover the effects of broader proteomic context on motif-mediated interactions, revealing diverse mechanisms of contextual control over EVH1 interactions and establishing that SLiMs cant be fully understood outside of their native context.

biochemistry↗

bioRxiv2018

Tertiary structural motif sequence statistics enable facile prediction and design of peptides that bind anti-apoptotic Bfl-1 and Mcl-1

Understanding the relationship between protein sequence and structure well enough to rationally design novel proteins or protein complexes is a longstanding goal in protein science. The Protein Data Bank (PDB) is a key resource for defining sequence-structure relationships that has supported the development of critical resources such as rotamer libraries and backbone torsional statistics that quantify the probabilities of protein sequences adopting different structures. Here, we show that well-defined, non-contiguous structural motifs (TERMs) in the PDB can also provide rich information useful for protein-peptide interaction prediction and design. Specifically, we show that it is possible to rapidly predict the binding energies of peptides to Bcl-2 family proteins as accurately as can be done with widely used structure-based tools, without explicit atomistic modeling. One benefit of a TERM-based approach is that prediction performance is less sensitive to the details of the input structure than are methods that evaluate energies using precise atomic coordinates. We show that protein design using TERM energies (dTERMen) can generate highly novel and diverse peptides to target anti-apoptotic proteins Bfl-1 and Mcl-1. 15 of 17 peptides designed using dTERMen bound tightly to their intended targets, and these peptides have just 15 - 38% sequence identity to any known native Bcl-2 family protein ligand. High-resolution structures of four designed peptides bound to their targets provided opportunities to analyze strengths and limitations of this approach. Dramatic success designing peptides using dTERMen, which comprised going from input structure to experimental validation of high-affinity binders in approximately one month, provides strong motivation for further developing TERM-based approaches to design.

biophysics↗