SEARCH · Search bioRxiv

Results for “Bioengineering”

Search indexed bioRxiv preprints in genomics, neuroscience, cell biology and bioinformatics. Read source abstracts and check manuscript versions; preprints are not peer reviewed.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 451 records · Page 25Linked to original sources

bioRxiv2025

EASY-CRISPR: a toolbox for high-throughput single-step custom genetic editing in bacteria

Targeted gene editing can be achieved using CRISPR/Cas9-assisted recombineering. However, high-efficiency editing requires careful optimization for each locus to be modified which can be tedious and time-consuming. In this work, we developed a simple, fast and cheap method for the Editing and Assembly of SYnthetic operons using CRISPR/Cas9-assisted recombineering (EASY-CRISPR) in Escherichia coli. Highly efficient editing of the different constitutive elements of the operons can be achieved by using a set of optimized guide RNAs and single- or double-stranded DNA repair templates carrying relatively short homology arms. This facilitates the construction of multiple genetic tools, including mutant libraries or reporter genes. EASY-CRISPR is also highly modular, as we provide alternative and complementary versions of the operon inserted in three loci which can be edited iteratively and easily combined. As a proof of concept, we report the construction of several fusions with reporter genes confirming known post-transcriptional regulation mechanisms and the construction of saturated and unbiased mutant libraries. In summary, the EASY-CRISPR system provides a flexible genomic expression platform that can be used both for the understanding of biological processes and as a tool for bioengineering applications.

microbiology↗

bioRxiv2025

Stacked mutations disrupting syringyl and p-coumaroylated lignin biosynthesis in rice result in lignin dominated by guaiacyl units: insights into grass-specific lignin monomer biosynthesis and polymerization mechanisms

O_LIThe aromatic composition of lignin significantly impacts the usability of lignocellulosic biomass. In eudicots, transgenic and mutant lines with elevated guaiacyl (G) or syringyl (S) lignin units have been successfully generated by manipulating the expression level of CONIFERALDEHYDE 5-HYDROXYLASE (CAld5H). However, this bioengineering approach has proven less effective in grasses, implicating the potential existence of a grass-specific alternative pathway for S lignin biosynthesis. C_LIO_LIThrough characterization of genome-edited rice mutants, we demonstrated that S lignin in rice can be virtually eliminated by disrupting genes encoding CAld5H along with p- COUMAROYL-COENZYME A:MONOLIGNOL TRANSFERASE (PMT), a grass-specific enzyme essential for the biosynthesis of monolignol p-coumarate conjugates. In contrast, individual mutations in either CAld5H or PMT genes resulted in incomplete elimination of S lignin. These findings provide strong evidence that rice possesses a CAld5H-independent pathway leading to the grass-specific monolignol p-coumarate conjugates. C_LIO_LIIn-depth structural characterizations of G-dominated lignins from rice and Arabidopsis mutants, natural gymnosperm pine, and G-type synthetic lignin revealed pronounced effects of lineage-dependent cell wall environments on the linkage patterns and molecular weight distributions of the resulting lignin polymers. C_LIO_LIOverall, our findings highlight previously overlooked lineage-specific lignin monomer biosynthesis and polymerization patterns in grasses. C_LI

plant biology↗

bioRxiv2025

Natural spider silk enhances mechanical performance of collagen scaffold under stretching conditions

Collagen is the most abundant protein in the extracellular matrix, crucial for wound healing and cell proliferation. While it holds promise as a scaffold for tendon, skin, and ligament reconstruction, collagens mechanical strength, particularly under stretch, is poor. Previous attempts to improve collagen strength involved blending it with silkworm or recombinant spider silk. In this study, for the first time, we evaluated whether collagen gel from fish skin could be strengthened by infusing it with native spider silk, specifically the major ampullate (MA) silk of Nephila pilipes, known for superior mechanical properties. MA silk was woven onto a frame, pressed into a PDMS platform, and then used to create a collagen scaffold. Youngs modulus of the infused collagen scaffold, subjected to either stretching or non-stretching treatments, was measured using AFM. After 24 hours of cyclic stretching, collagen infused with silk showed less fragility, higher Youngs modulus, and no bacterial growth. Immunohistochemical staining showed that after stretching, the thickness and architecture of the collagen gel infused with silk were maintained, and the fibers were reorganized in a more compact, aligned, and denser manner. Overall, collagen infused with native spider silk exhibited improved mechanical stability and stiffness under cyclic stretching, suggesting that this combination could serve as a robust matrix for bioengineering applications while preventing bacterial infiltration.

zoology↗

bioRxiv2025

An animal-free bioprocess to synthesize 3D human matrix scaffolds using mesenchymal stromal cells

Basement membrane is a specialized extracellular matrix (ECM) that compartmentalizes epithelial and endothelial tissues and provides essential structural and signaling cues for tissue organization. While fibrillar collagens (Col) of the interstitial matrix (e.g., Col I and Col III) are widely used in tissue modeling, the networking collagens that scaffold basement membrane, including human Col IV and Col VI, remain difficult to access. Commercial basement membrane surrogates such as Matrigel(R) are derived from murine tumors and are ill-defined, dilute, variable, and incompatible with animal-free biomanufacturing. Thus, there is a critical need for human-derived basement membrane matrices that are free of xenogenic contaminants and do not rely on animal agriculture. Here, we investigated whether human mesenchymal stromal cells (MSCs) could serve as a platform to produce self-assembling basement membrane components under chemically defined, xeno-free conditions. MSCs from placental, umbilical cord, bone marrow, and adipose tissues were cultured as three-dimensional spheroids and adherent multilayered sheets. Confocal imaging of whole-mount, decellularized spheroid matrices revealed complex networks of fibronectin and Col IV with topological and organizational features characteristic of native basement membrane. Perinatal MSCs produced distinct matrix architectures consisting of apical fibronectin sheets underlaid by continuous Col IV networks. Time-resolved imaging of umbilical cord MSC-derived matrix sheets demonstrated a reproducible sequence of basement membrane assembly that parallels developmental tissue organization. Together, these findings demonstrate that human MSCs, cultured entirely without animal-derived components, can synthesize functional basement membrane proteins that self-assemble into ordered, tissue-like scaffolds. This work establishes MSCs as a scalable, sustainable, and cruelty-free platform for the manufacturing of human basement membrane matrices for bioengineering and regenerative medicine applications.

cell biology↗

bioRxiv2025

A tRNA-gRNA multiplexing system for CRISPR genome editing in Marchantia polymorpha.

The liverwort Marchantia polymorpha is a widely used model organism for studying land plant biology, which has also proven to be a promising testbed for bioengineering. CRISPR/Cas9 technology has emerged as a transformative tool for precise genome modifications in M. polymorpha. However, a robust method for the simultaneous expression of multiple gRNAs, which is crucial for enhancing the efficiency and versatility of CRISPR/Cas9-based genome editing, has yet to be fully developed. In this study, we introduce an adaptation from the OpenPlant kit CRISPR/Cas9 tools, that facilitates expression of multiple gRNAs from a single transcript through incorporation of tRNA sequences. This approach significantly improves the efficiency and scalability of genome editing in M. polymorpha. Additionally, by combining this vector system with a simplified and optimized protocol for thallus transformation, we further streamline the generation of CRISPR/Cas9 mutants in M. polymorpha. The resulting gene- editing system offers a versatile, time-saving and straightforward tool for advancing functional genomics in M. polymorpha, enabling more comprehensive genetic modifications and genome engineering.

plant biology↗

bioRxiv2025

Interplay Between Intrinsically Disordered Proteins and Atomically Precise Gold Nanoclusters Modulates their Optical Properties

Understanding how structural and optical properties of metallic nanoclusters can be tuned by proteins is crucial for the use of these hybrid molecules in biomedical applications. The interaction of proteins with ultrasmall, atomically-precise gold nanoclusters (Au-NCs) has been mainly investigated in the context of structured proteins, while their behavior with intrinsically disordered proteins (IDPs) remains unexplored. This work examines the structural and optical properties of Au-NCs interacting with bioengineered IDPs containing up to three cysteines. We show that, by exploiting the conformational flexibility of cysteine-containing IDPs, we can anchor proteins to Au-NCs in a position-specific manner, leading to new bioconjugates with properties that differ from those of the individual components. We observed an up to 15-fold photoluminescence enhancement depending on the number of cysteines anchored. By combining mass spectrometry, small-angle X-ray scattering (SAXS), and computational modelling, the ensemble structures of nine bioconjugates with different stoichiometries were elucidated, indicating their overall compactness. Our results suggest that the interface between these atomically-precise species and the conformationally fluctuating protein is responsible for the optical properties of these nanobioconjugates. This research improves our understanding of Au-NC- protein interactions, paving the way to novel nano-molecular hybrid conjugates with tunable properties for bioimaging and therapeutic applications.

biochemistry↗

bioRxiv2025

Nuclear phenotypic changes in cells originating from migrating tenocytes in tendon explants

Tenocytes migrating from cultured rat tendon explants aggregate and gradually align in parallel rows with the tendons long axis as the explant period extends up to 12 days and collagen bundles become structurally remodeled. Tendon architecture and mechanical load are known to affect fibroblast phenotypes; thus, we studied cell cultures derived from tenocytes at early and later migration stages using image analysis and immunofluorescence for selected epigenetic markers to identify differences in nuclear characteristics across migration periods. Feulgen-DNA amounts and global DNA 5-methylcytosine and acetylated H3K9 distribution were not found to differ in the cells originating from tenocytes that migrated at the 8- and 12-day periods of tendon culture. However, light absorbance, fractal dimension, and margination features related to chromatin condensation and distribution in the Feulgen-stained cell nuclei differed during these periods. Accumulation of the H3K4me2 marker at the nuclear periphery in cells obtained after a longer explant culture period, when the margination feature reaches larger values, may represent the acquisition of an enhanced transcriptional memory that could affect future gene expression levels. Further assays are required to determine whether the present image analysis results correlate with the differential potential of these cells for extracellular matrix synthesis and practical bioengineering purposes.

cell biology↗

bioRxiv2025

Mechanistic insights into Enterocin C targeting the undecaprenyl phosphate recycling protein BacA

Enterococcus faecalis is an important opportunistic pathogen responsible for healthcare-associated infections. It is intrinsically resistant to various antibiotics, particularly to cephalosporins and vancomycin, creating an urgent need for alternative therapeutics. In this context, bacteriocins warrant investigations as a potential source of medical-use antibiotics. Herein, we demonstrate that Enterocin C, a class IIb two-peptide bacteriocin, specifically targets the membrane embedded undecaprenyl phosphate recycling protein BacA from enterococci as a cell surface receptor. Using biochemical and biophysical methods, supported by AlphaFold2 modelling and mutagenesis, we deciphered the EntCs molecular interaction pattern with its target, marking the first mechanistic insight of a two-peptide bacteriocin. The two peptides act cooperatively at nanomolar concentrations to interact with the outward-open catalytic pocket of BacA: the peptide EntC1 docks deeply into the catalytic site, inhibits BacAs enzymatic activity and enables the binding of peptide EntC2, eliciting membrane permeabilization, eventually leading to cell death. A comparative analysis with LcnG, a homologous bacteriocin, reveals a conserved interaction pattern, paving the way for bioengineering of these bacteriocins and developing tailored antimicrobial strategies.

biochemistry↗

bioRxiv2025

LlamaAffinity: A Predictive Antibody Antigen Binding Model Integrating Antibody Sequences with Llama3 Backbone Architecture

Antibody-facilitated immune responses are central to the bodys defense against pathogens, viruses, and other foreign invaders. The ability of antibodies to specifically bind and neutralize antigens is vital for maintaining immunity. Over the past few decades, bioengineering advancements have significantly accelerated therapeutic antibody development. These antibody-derived drugs have shown remarkable efficacy, particularly in treating Cancer, SARS-Cov-2, autoimmune disorders, and infectious diseases. Traditionally, experimental methods for affinity measurement have been time-consuming and expensive. With the realm of Artificial Intelligence, in silico medicine has revolutionized; recent developments in machine learning, particularly the use of large language models (LLMs) for representing antibodies, have opened up new avenues for AI-based designing and improving affinity prediction. Herein, we present an advanced antibody-antigen binding affinity prediction model (LlamaAffinity), leveraging an open-source Llama 3 backbone and antibody sequence data employed from the Observed Antibody Space (OAS) database. The proposed approach significantly improved over existing state-of-the-art (SOTA) approaches (AntiFormer, AntiBERTa, AntiBERTy) across multiple evaluation metrics. Specifically, the model achieved an accuracy of 0.9640, an F1-score of 0.9643, a precision of 0.9702, a recall of 0.9586, and an AUC-ROC of 0.9936. Moreover, this strategy unveiled higher computational efficiency, with a five-fold average cumulative training time of only 0.46 hours, significantly lower than previous studies. LlamaAffinity defines a new benchmark for antibody-antigen binding affinity prediction, achieving advanced performance in the immunotherapies and immunoinformatics field. Furthermore, it can effectively assess binding affinities following novel antibody design, accelerating the discovery and optimization of therapeutic candidates.

bioinformatics↗

bioRxiv2025

Architecture of the bundle-shaped phycobilisome from Gloeobacter violaceus

Cyanobacteria use soluble antenna megacomplexes, phycobilisomes (PBS), to maximize light-harvesting efficiency and small photoswitchable orange carotenoid proteins (OCPs) to down-regulate PBS in high light. Among known PBS morphologies, the one from the basal cyanobacterial genus Gloeobacter still lacks detailed structural characterization. Here, we present the cryo-EM structure of the G. violaceus PBS, a >10-MDa complex with a unique architecture consisting in 516 polypeptide chains totaling nearly 90,000 residues and harboring over 860 bilin chromophores. This unique PBS features diverging, conformationally mobile bundles of rods composed of stacked phycoerythrin and phycocyanin hexamers, stemming from a pentacylindrical allophycocyanin core belted by auxiliary phycocyanin hexamers. We show how two Gloeobacter-specific multidomain linker proteins, Glr1262 and Glr2806, maintain this bundle-shaped architecture and reveal its differential regulation via non-photochemical quenching by two OCP types of G. violaceus that recognize separate binding sites within the allophycocyanin core, including lateral cylinders absent in tricylindrical cores. Our findings provide the high-resolution structural insight into Gloeobacter PBS and its regulation, revealing divergent adaptations in early-branching cyanobacteria. The structure advances understanding of PBS diversity and evolution, offering a framework for bioengineering light-harvesting systems in synthetic biology and biotechnological applications.

biochemistry↗

bioRxiv2025

Association of E. coli Nissle with insect-parasitic nematode Steinernema hermaphroditum

Steinernema species are soil-dwelling and insect-parasitic nematodes that associate with symbiotic Xenorhabdus bacteria. During the infective juvenile (IJ) stage, Steinernema nematodes package species-specific Xenorhabdus bacteria in the anterior intestinal pockets. The nematodes can survive in the soil for months as they seek insect prey. The mechanisms of how these nematodes associate with environmental microbes other than their Xenorhabdus symbionts is barely known. Here, we report a new mechanism of E. coli Nissle (EcN) association with the nematode Steinernema hermaphroditum. We show that EcN cells are enclosed and lysed in at least four pairs of coelomocytes, suggesting these immune cells respond to bacterial invasion. During the IJ stage of nematode development, EcN cells localize to posterior intestinal vacuoles and enter the inter-cuticular space, where they proliferate, aggregate, then lyse. EcN cells expressed proteins in the cell lysates were maintained in the nematode cuticle over eight weeks in non-sterile soil. These observations suggest sequential steps of EcN colonization in the host nematodes involving an immune response that is distinctive from interactions with mutualistic symbiont. Our work establishes a novel framework of nematode-bacteria interaction with potential applications in environmental bioengineering.

microbiology↗

bioRxiv2025

Evaluation of natural killer cell tumor homing and effector function in response to CDK4/6 and AURKA inhibition in a melanoma tumor-on-a-chip platform

Natural killer (NK) cells have emerged as an important clinical tool cellular immunotherapy. Whereas immune checkpoint blockade (ICB) or chimeric antigen receptor (CAR) T-cell therapy (CAR-T) therapy have been adopted as a first line treatments in different malignancies, such as melanoma, these approaches do not work for all patients. T cells require proper antigen presentation on tumor cells for recognition and to carry out their corresponding cytotoxic functions. Deficiency of tumor antigens, or high variability in those present, make T cell-based CAR-T and ICB ineffective. By contrast, NK cells are not limited by antigen presentation deficiencies, offering a potential alternative approach, yet their efficacy can suffer from immunosuppressive signals. Herein, we sought to develop in vitro and on-chip platforms to identify strategies for enhance, rather than suppress, NK cell homing to tumor cells. We explored the use of inhibition of kinases such as CK4/6 and AURKA to induce tumor cell production of chemokines that NK cells migrate towards in aggressive melanoma models. We evaluated chemokine-aided NK cell migration-homing capabilities and their therapeutic efficacy and found that treatment of both melanoma cell line and patient-tumor constructs (PTCs) with CDK4/6 and AURKA generally resulted in improved NK cell homing to tumor cells and accompanying tumor cell killing. Interestingly, this chemokine-guided NK cell migration did not generate as effective outcomes in models using a mildly aggressive melanoma cell line. For our studies, we used 3D tumor constructs in both static Transwell models and then in a bioengineered NK cell-functionalized tumor-on-a-chip (NK-TOC) platform. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=83 SRC="FIGDIR/small/662476v1_ufig1.gif" ALT="Figure 1"> View larger version (20K): org.highwire.dtl.DTLVardef@17bf445org.highwire.dtl.DTLVardef@e1e1d9org.highwire.dtl.DTLVardef@1b29c2corg.highwire.dtl.DTLVardef@12b1f03_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2025

Efficient Data Collection for Establishing Practical Identifiability via Active Learning

Practical identifiability analysis (PIA) plays a crucial role in model development by determining whether available data are sufficient to yield reliable parameter estimates. In bioengineering applications, identifying the minimal experimental design that ensures parameter identifiability is essential in order to reduce cost, time, and resource consumption. In this paper, we introduce E-ALPIPE, a sequential active learning algorithm that recommends new data collection points most likely to establish practical identifiability given the current data, mathematical model and noise assumptions. We empirically evaluate E-ALPIPE against both a benchmark algorithm from the literature and random sampling over three synthetic experiments. Our results show that E-ALPIPE requires up to 50% fewer observations on average to achieve practical identifiability, compared to the strongest competitor, while producing comparable or narrower confidence intervals and more accurate point estimates of system dynamics. Author summaryScientists and engineers often use computer models to understand complex biological systems, such as how cells grow or how chemicals behave in a reactor. To make these models useful, they need to determine the correct values for various model parameters, based on data from experiments. However, experiments are usually expensive, time-consuming, and often produce noisy data. If the available data is insufficient, our confidence in the fitted model parameters may be low and thus the models predictions might be unreliable. This is known as a problem of practical identifiability. This paper introduces a new method called E-ALPIPE, which helps scientists decide which experiment is likely to produce the most useful information for improving their model. Using E-ALPIPE they can determine the model parameter values more efficiently, saving time and resources. We tested E-ALPIPE on several example problems and found that it could reach the same level of reliability using up to 50% fewer experiments than standard methods. It also produced more precise and accurate results. Overall, E-ALPIPE is a smarter, more efficient way to design experiments for tuning the parameters of computer models.

systems biology↗

bioRxiv2025

Assessing cellular metabolic dynamics with NAD(P)H fluorescence polarization imaging

Altered metabolism enables adaptive advantages for cancer, driving the need for improved methods for non-invasive long-term monitoring of cellular metabolism from organelle to population level. Here we present two-photon steady-state fluorescence polarization ratiometric microscopy (FPRM), a label-free imaging method that uses nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) autofluorescence as a functional readout of cellular metabolism. The method is simple to implement and operates an order of magnitude faster than the NAD(P)H-fluorescence lifetime imaging microscopy (FLIM) imaging modality, reducing cytotoxic stress while providing long-term monitoring capacity. FPRM enables high-resolution dynamic tracking of NAD(P)H signals with subcellular details and we have established a set of instrument-independent ratiometric parameters that correlates NAD(P)H signals with metabolic status during pharmaceutical and environmental perturbations. We further integrated FPRM readouts with other parameters such as cell shape and migration on 2D and 3D collagen matrices, demonstrating the techniques versatility across bioengineered platforms for cancer metabolism research.

biophysics↗

bioRxiv2025

Cytoplasmic Abundant Heat-Soluble Proteins from Tardigrades Protect Synthetic Cells Under Stress

Cytoplasmic abundant heat-soluble (CAHS) proteins, a stress-responsive intrinsically disordered protein from tardigrades, have been discovered to form gel-like networks providing structural support during dehydration, thus enabling anhydrobiosis. However, the mechanism by which CAHS proteins protect the dehydrating cellular membrane remains enigmatic. Using giant unilamellar vesicles (GUVs) as a model membrane system, we show that encapsulated CAHS12 undergoes a reversible structural transformation that reinforces membrane integrity and preserves encapsulated components, mimicking natural anhydrobiosis. CAHS12-containing GUVs demonstrated stability for weeks and mechanical robustness under dehydration, elevated temperature, and osmotic stresses. Molecular simulations suggest that CAHS12 forms a filamentous network within the vesicle lumen that mitigates membrane collapse and preserves compartmental architecture. Synthetic cells with cell-free transcription-translation capabilities withstand desiccation and recover biochemical activities, akin to the tun state of the tardigrade. This discovery opens up synthetic cell applications in bioengineering, cold-chain-independent biomanufacturing, and adaptive biointerfaces.

synthetic biology↗

bioRxiv2025

Stable Synthetic Organelles from Aqueous Two-Phase Systemswith Access to the Cell Translation Machinery

Cells compartmentalize vital processes in membrane-less organelles to gain spatiotemporal control of metabolism, signaling, and for protection under stress. While such compartments can be manipulated or even de novo designed with genetic engineering of cells, the transfer and operation of exogenous synthetic compartments for intracellular engineering is challenged by developing pathways for implantation into the cell, stability issues, toxicity, ability to maintain compartmentalization, and access to cellular machinery. Here we introduce dextran-lipid droplets as versatile exogenous synthetic compartments, that are readily uptaken by model cancer and immune cells and are stable inside cells for days. Furthermore, the droplets can encapsulate nucleic acids with high efficiency, are non-toxic and endosomal escape occurs when formulated with ionizable lipids, as shown by reporter protein translation of an mRNA hosted inside the organelles. We propose that such synthetic microreactors implanted into the cells will become an important bioengineering tool to incorporate more complex and bioorthogonal molecular systems into cells.

synthetic biology↗

bioRxiv2025

Evaluation of Bacterial Microcompartment Cofactor Recycling and Permeability with a Model Guided In Vitro Assay

Biomanufacturing is a promising strategy for sustainable chemical production. However, challenges such as cofactor competition and low pathway flux prevent competitive titers. Some bacteria address these challenges by encapsulating metabolic pathways in bacterial microcompartments (MCPs), many of which contain dedicated cofactor recycling enzymes. We sought to determine how pathway cofactor recycling and intermediate sequestration in MCPs benefit pathway performance using an in vitro assay and kinetic model of the 1,2-propanediol utilization (Pdu) system. Guided by model simulations, we performed experimental design to characterize permeability, a key and difficult-to-measure property of MCPs. Using our model and measurements of metabolite concentrations over time, we estimate MCP permeability values in the range of 10-5 cm/s. We also demonstrated that NAD+/NADH recycling in the Pdu MCP benefits increased pathway flux. This study integrates experiments and systems modeling to advance our understanding of why pathways are encapsulated and to inform bioengineering applications.

biochemistry↗

bioRxiv2025

The genetic architecture of the human bZIP family

Generative biology holds the promise to transform our ability to design and understand living systems by creating novel proteins, pathways, and organisms with tailored functions that address challenges in medicine, sustainability, and technology. However, training generative models requires large quantities of data that captures the genetic architecture of protein function in all its complexity, but these are currently scarce. Here, we systematically mutagenized all 54 human basic-leucine zipper (bZIP) domains and quantified their interactions with each other using bindingPCA, a quantitative deep mutational scanning assay. This resulted in [~]2 million interaction measurements, capturing the effect of all single amino acid substitutions at each of the 35 interfacial positions. We found that mutation effects are largely additive in the vicinity of each wild-type bZIPs, but diverge across the family, indicating strong context dependency. A global additive thermodynamic model provided moderate prediction of mutation effects, while individual models per bZIP achieved higher performance, supporting a model of local simplicity and global complexity. Our results therefore suggest that the genetic architecture of protein function is more complex than previously anticipated, which could hinder predictability. However, a convolutional neural network trained on this dataset could accurately predict binding scores from sequence alone. Furthermore, the model enabled the design of synthetic bZIPs with high target specificity, demonstrating practical applicability for bioengineering purposes. Our study shows that capturing family-wide diversity is essential to reveal context dependencies and train accurate quantitative models of protein-protein interactions.

biophysics↗