Biology subjects

Steel, H.

Publications and source records attributed to Steel, H..

At least 19 recordsLinked to original sources

bioRxiv2026

Machine learning-assisted directed evolution yields dramatic improvement on novel AAV engineering task

Directed evolution enables the discovery of protein mutants with improved fitness through iterative rounds of selection and has been widely applied to adeno-associated virus (AAV) capsid engineering. Machine learning (ML) can augment this process by prioritising mutants for experimental validation, but whether its benefits outweigh the added cost of ML-designed library construction remains unclear. Here, we address this question by considering improvements in manufacturing efficiency of AAV capsids in the context of exosomal encapsulation, which is a technique for improving AAV immunogenicity. We generated a directed evolution dataset comprising 53,974 mutants across three rounds of selection and an independent assessment dataset of 472 ML-designed mutants with detailed profiling. Directed evolution alone yielded a 3-fold improvement, while ML-assisted directed evolution yielded a 41-fold improvement in manufacturing efficiency, demonstrating that ML-assisted directed evolution outperforms directed evolution alone. We show that data quality - specifically the number of selection rounds and the statistical power of mutant counts - has a greater impact on model performance than model architecture. Finally, we provide practical guidance on experimental design and implementation of ML tuning strategies that can augment model performance in protein engineering applications.

bioengineering↗

bioRxiv2026

Quantitative Engineering and Investigation of Synthetic Sponge RNAs in E. coli

Sponge RNAs (spRNAs) play an important regulatory role in bacterial small RNA (sRNA) networks, but their engineering and quantitative systems-level properties are unexplored. Here, we design, build, and quantitatively characterise synthetic spRNA-based gene circuits in E. coli. We establish multiple design strategies for synthetic spRNAs, engineering the first synthetic spRNAs. We show that these synthetic spRNAs can reversibly de-repress sRNA-regulated gene expression, demonstrate tuneable control of gene expression, and extend these designs to multi-target regulation. Through the use of time-resolved continuous-culture characterisation in Chi.Bio together with absolute fluorescent protein quantification, we generated a quantitative dynamical dataset for model fitting and mechanistic analysis. Sequential model development showed that recapitulating the observed circuit dynamics required incorporation of Hfq-mediated resource competition, often overlooked in models of sRNA-based synthetic gene circuits. The extended model captured promoter, sRNA, and sponge circuit behaviour and was used to investigate quantitative properties of spRNA-mediated regulation, the first such quantitative investigation of spRNA-based regulation. Model-based quantitative investigations further suggest that spRNAs can tune response functions, modulate thresholds and leakiness, alter response times, improve disturbance rejection in some regimes, increase effective specificity, and buffer regulatory output against sRNA mutation. Together, these results establish synthetic spRNAs as a new post-transcriptional tool for bacterial synthetic biology and provide a quantitative framework for understanding natural and engineered spRNA-mediated regulation.

synthetic biology↗

bioRxiv2026

A Toolbox for Biomanufacturing of Functionalised PHA Nanoparticles with C. necator

Microbes have the potential to manufacture plastics from sustainable feedstocks while enabling novel material properties and functions that are not easily accessible through conventional chemical synthesis. Realising this potential requires a comprehensive genetic and process engineering framework that spans chassis and bioprocess optimisation, polymer property control, and downstream functionalisation. Here we develop such a platform in Cupriavidus necator, with a focus on high-value polyhydroxyalkanoate (PHA) nanoparticles. To this end we first optimise the transformation protocol for the organism. Next, we create a library of PhaC synthase variants from C. necator, Aeromonas caviae and Brevundimonas sp. in a {Delta}phaC background, demonstrating that they allow customisation of the material properties of produced PHA particles. Our results combine data from Flow cytometry, Transmission Electron Microscopy (TEM), Fourier Transform InfraRed Spectroscopy (FTIR), and Differential Scanning Calorimetry (DSC) to show that it is possible to generate materials ranging from highly crystalline PHAs to softer P(3HB-co-3HHx) copolymers and that an A. caviae PhaC variant can double the yield of large PHA granules. To improve bioprocess sustainability, we coupled C. necator with B. subtilis in sucrose-fed co-cultures, using tetracycline tolerance differences and inoculation ratios to enhance PHA production from inexpensive, sugar-rich feedstocks. Finally, we add function to the produced PHA nanoparticles by using the molecular protein-fusion technology SpyTag-SpyCatcher, showing it is possible to efficiently capture SpyCatcher-GFP on PHA granules as a proof of concept for PHAs use as a customisable bio-based nanoparticle. Together, our work offers an innovation to produce bio-PHA nanoparticles in a customisable way, with potential applications in sustainable biomanufacturing, biosensing, drug delivery and future bioremediation technologies.

synthetic biology↗

bioRxiv2025

Closing the Loop on Phage-bacteria Coevolution

Bacteria and their viruses, bacteriophages (phages), continually coevolve in nature. In contrast, laboratory-based coevolution experiments usually last less than a month before becoming dormant or extinct as one species is outcompeted by the other. Consequently, there is a poor understanding of phage-bacteria coevolution and hence the long-term efficacy of bacteriophage therapies (an approach to tackling antimicrobial resistance). We propose a novel approach to coevolution experiments that would address this challenge: instead of open-loop resource-constrained cultures, we develop a closed-loop control approach to stabilise the typically unstable or oscillatory phage-bacteria population dynamics. Achieving this requires the control system to compensate for delays in phage incubation and respond to an evolving system, while only measuring bacterial density. To this end, we develop a model of phage-bacteria dynamics, prototype delay-compensating predictive control strategies, and demonstrate a measurement-aware state observer. Overall, this approach shows the ability to stabilise coevolution, avoiding the common outcomes of unstable dynamics or winner-takes-all competition.

evolutionary biology↗

bioRxiv2025

Engineered AAV Capsids with Enhanced Extracellular Vesicle Loading via Rational Design and Directed Evolution

Overcoming immune responses and ensuring efficient delivery remain key obstacles to advancing adeno-associated virus (AAV)-based genetic medicines into the clinic. Encapsulating AAVs within naturally secreted extracellular vesicles (EV-AAV) can shield them from neutralising antibodies and increase transduction efficiency; however, wild-type AAV capsids are incorporated into EVs inefficiently, with the percentage of total AAV genomes in the EV-AAV fraction ranging from 0.5% to 12%, depending on AAV serotype. Here, we establish a high-throughput screening platform that links AAV capsid genotype to EV encapsulation phenotype to improve encapsulation efficiency. We combined rational design with directed evolution to engineer AAV9 variants that preferentially sort into EVs. A library of 392,550 AAV9 capsid mutants was generated by inserting Late (L) domains, established EV-sorting motifs, and fully random amino acid inserts in a surface-exposed region of the AAV9 capsid protein. Screening data were analysed using a statistical model-assisted protein selection (SMAPS) workflow - developed as a novel quantitative modelling and selection strategy for protein engineering challenges of this kind. SMAPS functions as a digital twin of the protein variants in our high-throughput screen, allowing us to monitor uncertainty in copy numbers over multiple steps in the experimental workflow. Using SMAPS, we identified five lead hits, which showed up to threefold increased EV-AAV yield compared to wild-type, without altering particle size. Moreover, in vitro transduction experiments demonstrated that the lead AAV9 capsid exhibits substantially higher transduction efficiency compared to wild-type AAV9 and retained functional activity in the presence of neutralising anti-AAV9 antibodies. Collectively, our study: (i) delivers a scalable high-throughput screen for EV-AAV engineering, (ii) introduces the fields first statistical model-guided quality-control framework for directed evolution, and (iii) provides best-in-class capsids for EV encapsulation.

synthetic biology↗

bioRxiv2025

Microscopic PhotoSelection (MiPS) of single cells in mother machine microfluidic devices

Techniques for selecting or sorting single cells within large populations of genetic variants are central to synthetic biology and biotechnology. Widely-used methods such as Fluorescence Activated Cell Sorting (FACS) enable rapid processing of large libraries, but are restricted to low-dimensional measurements taken at a single time point. As a result, sorting based on dynamic or multi-trait phenotypes--such as transient properties and properties that occur on fast timescales or in response to dynamic actuating signals--remains fundamentally challenging. Here we introduce Microscopic PhotoSelection (MiPS) which employs an automated robotic platform for single-cell selection based on multiple dynamic criteria, directly on microfluidic mother machine devices. The system couples long-term single-cell imaging with real-time analysis and selective optical targeting, allowing fully automated enrichment using high-intensity UV light or alternative wavelengths with the addition of photosensitisers. By targeting many cells in parallel, our platform overcomes throughput limitations of existing microfluidic-based selection technologies such as optical tweezers or droplet-based methods and provides a direct approach to select cells based on time-resolved, multi-trait phenotypes. We demonstrate the ability to perform in vivo selection and outline how iterative, feedback-based selection strategies can refine enrichment across multiple rounds. Taken together, our work establishes a high-throughput selection framework integrated into microfluidic devices, enabling applications in directed evolution, biosensor optimisation, circuit engineering, and diagnostics, where selection based on dynamic, multi-trait phenotypes is essential.

bioengineering↗

bioRxiv2025

Sequence-free landscape inference for directed evolution

Directed evolution is a method for engineering biological systems or components, such as proteins, wherein desired traits are optimised through iterative rounds of mutagenesis and selection of fit variants. The process of protein directed evolution can be envisaged as navigation over high-dimensional landscapes with numerous local maxima. The performance of any strategy in navigating such a landscape is dependent on the ruggedness of that landscape. However, this information is generally unavailable at the outset of an experiment, and cannot currently be computed using analytical methods. Here we propose SLIDE, Sequence-free Landscape Inference for Directed Evolution, which consists of two parts. First, SLIDE provides an estimation for landscape ruggedness from a mutating population using only population-level phenotypic data and an estimation of mutation rate. Ruggedness information in itself is valuable in protein design, for instance in predicting evolutionary stability. Second, SLIDE offers a framework for using the estimated ruggedness metric to select high-performing parameters for directed evolution control. Using theoretical NK landscapes and four real-world protein fitness landscapes, we demonstrate improvement upon the performance of standard selection strategies, particularly on rugged landscapes, using a pipeline that could also be combined with emerging AI-based methods for driving direction evolution.

synthetic biology↗

bioRxiv2025

Generative design of synthetic gene circuits for functional and evolutionary properties

1AbstractsIn the past decades, a wide suite of design tools for biological systems have been developed, but using these to create biotechnologies that achieve reliable and predictable behaviour remains challenging. Modelling approaches have enabled researchers to traverse the vast search space of genetic circuits more efficiently, while machine learning has proven useful for designing parts and predicting their function or evolutionary properties. Generative algorithms have the potential to leverage these features to design entire genetic circuits from the sequence level, but have only recently begun to be applied to synthetic biological applications. Here, we show that even simple generative models like the conditional variational autoencoder (CVAE) can produce novel genetic circuits that match complex dynamic functions such as signal adaptation. Using in silico RNA simulation, we construct a dataset of RNA sequences and convert them to circuits via RNA interaction predictors, allowing us to estimate functional features alongside evolutionary stability and interpret model-learned features. Our model generates diverse distributions of circuits that match their target adaptation specification well, even when limited to small training data sets. Structures in the embedding space correspond to motifs previously identified as crucial for adaptation and reflect the design rules for adaptable circuits. Framing adaptation as a single design objective outperforms other input representations, reflecting the importance of choosing the correct data encoding for generating genetic circuits. Finally, we show that functional and evolutionary properties can be prompted simultaneously, providing a proof-of-concept for the combined design of phenotype and evotype.

synthetic biology↗

bioRxiv2025

Automatic feedback control for resource-aware characterisation of genetic circuits

Many applications of engineered cells are enabled by genetic circuits - networks of genes regulating each other to process signals. Complex circuits are built by combining standardised modular components with different functions. Nonetheless, genes in a cell compete for the same limited pool of cellular resources, causing unintended interactions that violate modularity. Thus, circuit components behave differently when combined versus when observed in isolation, which can compromise a biotechnologys predictability and reliability. To forecast steady-state interactions between modules, experimental protocols for characterising their resource competition properties have been proposed. However, they rely on open-loop batch culture techniques in which dynamic control signals cannot be applied to cells. Consequently, these experimental methods have limited predictive power, as they may fail to capture all possible steady states, such as repelling equilibria that would not be approached by a system without external forcing. In contrast, we propose a novel, comprehensive protocol for characterising the resource-dependence of genetic modules performance. Based on the control-based continuation technique, it captures both stable and unstable steady states by applying stabilising cybergenetic feedback with an automated cell culturing platform. Using several models with different degrees of complexity, we simulate applying our pipeline to a self-activating genetic switch. This case study illustrates how informative characterisation of a genetic module with automatic feedback control enables reliable forecasting of its performance when combined with any other circuit component. Hence, our protocol promises to restore predictability to the design of genetic circuits from standardised components.

synthetic biology↗

bioRxiv2024

Quantum Correlations in Engineered Magneto-Sensitive Fluorescent Proteins Enables Multi-Modal Sensing in Living Cells

Quantum mechanical phenomena have been identified as fundamentally significant to an increasing number of biological processes. Simultaneously, quantum sensing is emerging as a cutting-edge technology for diverse applications across materials and biological science. However, until recently, biological based candidates for quantum sensors have been limited to in vitro systems, were prone to light induced degradation, and the experimental setups involved are typically not amenable to high-throughput study as would enable further engineering e.g. via directed evolution. We recently created a new class of magneto-sensitive fluorescent proteins (MFPs), which we show overcome these challenges and represent a new form of engineered biological quantum sensors that function both at physiological conditions and in living cells. Through directed evolution, we demonstrate the possibility of engineering these proteins to alter properties of their response to magnetic fields and radio frequencies. These effects are explained in terms of the radical pair mechanism (RPM), involving the protein backbone and a bound flavin cofactor. Using this engineered system we demonstrate the first observation of a fluorescent protein exhibiting Optically Detected Magnetic Resonance (ODMR) in living bacterial cells at room temperature, at sufficiently high signal-to-noise to be detected in a single cell. These magnetic resonance and magnetic field effects measured via fluorescence enable novel technologies; examples we demonstrate include spatial localisation of fluorescence signals using gradient fields (i.e. Magnetic Resonance Imaging (MRI) using a genetically encoded probe), sensing of the molecular microenvironment, multiplexing of bio-imaging, and lock-in detection, overcoming typical fluorescence imaging challenges of light scattering and autofluorescence. Taken together, our results represent a new range of sensing modalities for engineered biological systems, based on and designed around understanding the quantum mechanical properties of MFPs.

bioengineering↗

bioRxiv2024

Streamlined and efficient genome editing in Cupriavidus necator H16 using an optimised SIBR-Cas system

Cupriavidus necator H16 is a promising microbial platform strain for CO2 valorisation. While C. necator is amenable to genome editing, existing tools are often inefficient or rely on lengthy protocols, hindering its rapid transition to industrial applications. In this study, we simplified and accelerated the genome editing pipeline for C. necator by harnessing the Self-splicing Intron-Based Riboswitch (SIBR) system. We used SIBR to tightly control and delay Cas9-based counterselection, achieving >80% editing efficiency at two genomic loci within 48 hours after electroporation. To further increase the versatility of the genome editing toolbox, we upgraded SIBR to SIBR2.0 and used it to regulate the expression of Cas12a. SIBR2.0-Cas12a could mediate gene deletion in C. necator with [~]70% editing efficiency. Overall, we streamlined the genome editing pipeline for C. necator, facilitating its potential role in the transition to a bio-based economy.

bioengineering↗

bioRxiv2024

Model-guided gene circuit design for engineering genetically stable cell populations in diverse applications

Maintaining engineered cell populations genetic stability is a key challenge in synthetic biology. Synthetic genetic constructs compete with a host cells native genes for expression resources, burdening the cell and impairing its growth. This creates a selective pressure favouring mutations which alleviate this growth defect by removing synthetic gene expression. Non-functional mutants thus spread in cell populations, eventually making them lose engineered functions. Past work has attempted to limit mutation spread by coupling synthetic gene expression to survival. However, these approaches are highly context-dependent and must be tailor-made for each particular synthetic gene circuit to be retained. In contrast, we develop and analyse a biomolecular controller which depresses mutant cell growth independently of the mutated synthetic genes identity. Modelling shows how our design can be deployed alongside various synthetic circuits without any re-engineering of its genetic components, outperforming extant gene-specific mutation spread mitigation strategies. Our controllers performance is evaluated using a novel simulation approach which leverages resource-aware cell modelling to directly link a circuits design parameters to its population-level behaviour. Our designs adaptability promises to mitigate mutation spread in an expanded range of applications, whilst our analyses provide a blueprint for using resource-aware cell models in circuit design.

synthetic biology↗

bioRxiv2024

Cybernetic control of a natural microbial co-culture

A key obstacle in the widespread application of microbial co-cultures in bioprocesses is their compositional instability, as faster-growing species outcompete and dominate the culture. While several synthetic biology approaches have demonstrated control over co-culture composition, there has been an increased interest in computer-based cybernetic control approaches that can offload burdensome genetic control circuitry to computers and enable dynamic control and real-time noise rejection. This work extends that approach, demonstrating a cybernetic control method that is not reliant on any genetic engineering, instead interfacing cells with computers by exploiting their natural characteristics to measure and actuate the composition. We apply this to a Pseudomonas putida (P. putida) and Escherichia coli (E. coli) co-culture grown in Chi.Bio bioreactors, first showing how composition estimates calculated from different bioreactor measurements can be combined with a system model using an extended Kalman filter to generate accurate estimates of a noisy system. We also demonstrate that because the species have different optimal temperature niches, adjusting the temperature of the culture can drive the composition in either direction. By using a proportional-integral control algorithm to calculate the temperature that would bring the measured composition towards the desired composition, we are able to track dynamic references and stabilised the co-culture for 7 days ([~]250 generations), with the experiment ending before the cells could adapt out of the control. This cybernetic framework is broadly applicable, with different microbes unique features and specific growth niches enabling robust control over diverse co-cultures.

systems biology↗

bioRxiv2024

Optimised control of adaptive evolution with competing selective pressures

The development of methods to understand and control the population dynamics of microbial evolution remains an outstanding question in synthetic biology and biotechnology more broadly. Due to the stochastic nature of evolution, its limited observability, and complex intra-population dynamics, this presents a significant challenge. In this paper, we explore techniques to control the evolutionary dynamics of a population, based on manipulation of one or two orthogonal selective pressures, which may in turn be coupled to mutagenesis. Our approach builds on past research in evolutionary biology that developed frameworks to study intra-population variant competition during asexual adaptive processes (i.e. clonal interference). Extending this theory, we design optimal control strategies for one (or more) selective pressures that can be used to maximise the rate of adaptation across a population as a whole. We introduce a theoretical modelling framework for this process, which we support with both simulations and preliminary experimental data, providing a concrete basis for emerging control approaches to directed evolution and evolution-aware design.

evolutionary biology↗

bioRxiv2024

Combining positive and negative regulation for modular and robust biomolecular control architectures

Engineered biotechnologies are powered by synthetic gene regulation and control systems, known as genetic circuits, which must be modular and robust to disturbances if they are to perform reliably. An emerging family of regulatory mechanisms is mediated by clustered interspaced palindromic repeats (CRISPR) that can both interfere with (downregulate) or activate (upregulate) a given genes expression. However, all CRSIPR regulation relies on a shared resource pool of dCas9 proteins. Hence, a circuits components can indirectly affect one another via resource competition - even without any intended interactions between them - which compromises the modularity of synthetic biological designs. Using a resourceaware model of CRISPR regulation, we find that circuit modules which simultaneously subject a gene to CRISPR interference and activation are rendered robust to resource competition crosstalk. Evaluating this architectures simulated performance, we identify the scenarios where it can be advantageous over the extant resource competition mitigation strategies. We then consider different feedback architectures to demonstrate that combining opposite regulatory interactions overcomes the trade-off in robustness to perturbations of different nature. The motif of combined positive and negative regulation may therefore give rise to more robust and modular biomolecular controllers, as well as hint at the characteristics of natural systems that possess it.

synthetic biology↗

bioRxiv2024

Optimisation strategies for directed evolution without sequencing

Directed evolution can enable engineering of biological systems with minimal knowledge of their underlying sequence-to-function relationships. A typical directed evolution process consists of iterative rounds of mutagenesis and selection that are designed to steer changes in a biological system (e.g. a protein) towards some functional goal. Much work has been done, particularly leveraging advancements in machine learning, to optimise the process of directed evolution. Many of these methods, however, require DNA sequencing and synthesis, making them resource-intensive and incompatible with developments in targeted in vivo mutagenesis. Operating within the experimental constraints of established sorting-based directed evolution techniques (e.g. Fluorescence-Activated Cell Sorting, FACS), we explore approaches for optimisation of directed evolution that do not require sequencing information. We then expand our methods to the context of emerging experimental techniques in directed evolution, which allow for single-cell selection based on fitness objectives defined from any combination measurable traits. Finally, we validate the developed selection strategies on the GB1 empirical landscape, demonstrating that they can lead to up to a 7.5 fold increase in the probability of attaining the global fitness peak. Author summaryThe standard approach to sorting-based selection in directed evolution is to take forward only the top-performing variants from each generation of a single population. In this work, we begin to explore alternative selection strategies within a simulated directed evolution framework. We propose "selection functions", which allow us to tune the balance of exploration and exploitation of a fitness landscape, and we demonstrate that splitting a population into sub-populations can improve both the likelihood and magnitude of a successful outcome. We also propose strategies to leverage emerging selection methods that can implement single-cell selection based on any combination of measurable traits. We validate our optimised directed evolution approaches on the empirical fitness landscape of the GB1 immunoglobulin protein.

bioengineering↗

bioRxiv2024

Integration of pH control into Chi.Bio reactors and demonstration with small-scale enzymatic poly(ethylene terephthalate) hydrolysis

Small-scale bioreactors that are affordable and accessible would be of major benefit to the research community. In previous work, an open-source, automated bioreactor system was designed to operate up to the 30 mL scale with online optical monitoring, stirring, and temperature control, and this system, dubbed Chi.Bio, is now commercially available at a cost that is typically 1-2 orders of magnitude less than commercial bioreactors. In this work, we further expand the capabilities of the Chi.Bio system by enabling continuous pH monitoring and control through hardware and software modifications. For hardware modifications, we sourced low-cost, commercial pH circuits and made straightforward modifications to the Chi.Bio head plate to enable continuous pH monitoring. For software integration, we introduced closed-loop feedback control of the pH measured inside the Chi.Bio reactors and integrated a pH-control module into the existing Chi.Bio user interface. We demonstrated the utility of pH control through the small-scale depolymerization of the synthetic polyester, poly(ethylene terephthalate) (PET), using a benchmark cutinase enzyme, and compared this to 250 mL bioreactor hydrolysis reactions. The results in terms of PET conversion and rate, measured both by base addition and product release profiles, are statistically equivalent, with the Chi.Bio system allowing for a 20-fold reduction of purified enzyme required relative to the 250 mL bioreactor setup. Through inexpensive modifications, the ability to conduct pH control in Chi.Bio reactors widens the potential slate of biochemical reactions and biological cultivations for study in this system, and may also be adapted for use in other bioreactor platforms.

bioengineering↗

bioRxiv2024

Learning-Based Estimation of Fitness Landscape Ruggedness for Directed Evolution

Directed evolution is a method for engineering biological systems or components, such as proteins, wherein desired traits are optimised through iterative rounds of mutagenesis and selection of fit variants. The process of protein directed evolution can be envisaged as navigation over high-dimensional landscapes with numerous local maxima, mapping every possible variant of a protein to its fitness. The performance of any strategy in navigating such a landscape is dependent on several parameters, including its ruggedness. However, this information is generally unavailable at the outset of an experiment, and cannot be computed using analytical methods. Here we propose a learning-based method for estimating landscape ruggedness from a mutating population, using only population average performance data. This method uses a short period of exploration at the beginning of an experiment to predict the ruggedness, subsequently guiding the choice of high-performing parameters for directed evolution control. We then simulate this approach on two real-world protein fitness landscapes, demonstrating an improvement upon the performance of standard strategies, particularly on rugged landscapes. In addition to improving the overall outcomes of directed evolution, this method has the advantage of being readily deployable in laboratory settings, even in configurations that exclusively capture average population measures. Given the rapidly expanding application space of engineered proteins, the products of improved directed evolution are relevant in medicine, agriculture and manufacturing.

bioengineering↗