Bridging the gap between mechanistic biological models and machine learning surrogates.

Mechanistic models have been used for centuries to describe complex interconnected processes, including biological ones. As the scope of these models has widened, so have their computational demands. This complexity can limit their suitability when running many simulations or when real-time results are required. Surrogate machine learning (ML) models can be used to approximate the behaviour of complex mechanistic models, and once built, their computational demands are several orders of magnitude lower. This paper provides an overview of the relevant literature, both from an applicability and a theoretical perspective. For the latter, the paper focuses on the design and training of the underlying ML models. Application-wise, we show how ML surrogates have been used to approximate different mechanistic models. We present a perspective on how these approaches can be applied to models representing biological processes with potential industrial applications (e.g., metabolism and whole-cell modelling) and show why surrogate ML models may hold the key to making the simulation of complex biological systems possible using a typical desktop computer.

A proteomic approach identifies many novel palmitoylated proteins in Arabidopsis.

S-acylation (palmitoylation) is a poorly understood post-translational modification of proteins involving the addition of acyl lipids to cysteine residues. S-acylation promotes the association of proteins with membranes and influences protein stability, microdomain partitioning, membrane targeting and activation state. No consensus motif for S-acylation exists and it therefore requires empirical identification. Here, we describe a biotin switch isobaric tagging for relative and absolute quantification (iTRAQ)-based method to identify S-acylated proteins from Arabidopsis. We use these data to predict and confirm S-acylation of proteins not in our dataset. We identified c. 600 putative S-acylated proteins affecting diverse cellular processes. These included proteins involved in pathogen perception and response, mitogen-activated protein kinases (MAPKs), leucine-rich repeat receptor-like kinases (LRR-RLKs) and RLK superfamily members, integral membrane transporters, ATPases, soluble N-ethylmaleimide-sensitive factor-activating protein receptors (SNAREs) and heterotrimeric G-proteins. The prediction of S-acylation of related proteins was demonstrated by the identification and confirmation of S-acylation sites within the SNARE and LRR-RLK families. We showed that S-acylation of the LRR-RLK FLS2 is required for a full response to elicitation by the flagellin derived peptide flg22, but is not required for localization to the plasma membrane. Arabidopsis contains many more S-acylated proteins than previously thought. These data can be used to identify S-acylation sites in related proteins. We also demonstrated that S-acylation is required for full LRR-RLK function.

Pollen-tube tip growth requires a balance of lateral propagation and global inhibition of Rho-family GTPase activity.

Rapid tip growth allows for efficient development of highly elongated cells (e.g. neuronal axons, fungal hyphae and pollen tubes) and requires an elaborate spatiotemporal regulation of the growing region. Here, we use the pollen tube as a model to investigate the mechanism regulating the growing region. ROPs (Rho-related GTPases from plants) are essential for pollen tip growth and display oscillatory activity changes in the apical plasma membrane (PM). By manipulating the ROP activity level, we showed that the PM distribution of ROP activity as an apical cap determines the tip growth region and that efficient tip growth requires an optimum level of the apical ROP1 activity. Excessive ROP activation induced the enlargement of the tip growth region, causing growth depolarization and reduced tube elongation. Time-lapse analysis suggests that the apical ROP1 cap is generated by lateral propagation of a localized ROP activity. Subcellular localization and gain- and loss-of-function analyses suggest that RhoGDI- and RhoGAP-mediated global inhibition limits the lateral propagation of apical ROP1 activity. We propose that the balance between the lateral propagation and the global inhibition maintains an optimal apical ROP1 cap and generates the apical ROP1 activity oscillation required for efficient pollen-tube elongation.

Design and Assembly of Multilevel Transcriptional and Translational Regulators for Stringent Control of Gene Expression.

Precise control of gene expression is crucial when reprogramming the behavior of living cells. However, common inducible systems often lack the ability to stringently control gene expression due to the use of a single type of regulator that can be susceptible to unavoidable biomolecular fluctuations. In contrast, multilevel controllers (MLCs) employ several forms of regulation simultaneously to overcome this issue, ensuring a reduced basal expression while minimally affecting the maximum induced expression level that can be achieved. Here, we show how our publicly available genetic toolkit can be used to simplify the assembly of MLCs for the stringent control of gene expression. We demonstrate how new compatible parts can be designed and explain the rapid end-to-end assembly procedure.

Towards an engineering theory of evolution.

Biological technologies are fundamentally unlike any other because biology evolves. Bioengineering therefore requires novel design methodologies with evolution at their core. Knowledge about evolution is currently applied to the design of biosystems ad hoc. Unless we have an engineering theory of evolution, we will neither be able to meet evolution's potential as an engineering tool, nor understand or limit its unintended consequences for our biological designs. Here, we propose the evotype as a helpful concept for engineering the evolutionary potential of biosystems, or other self-adaptive technologies, potentially beyond the realm of biology.

Testing Theoretical Minimal Genomes Using Whole-Cell Models.

The minimal gene set for life has often been theorized, with at least ten produced for Mycoplasma genitalium (M. genitalium). Due to the difficulty of using M. genitalium in the lab, combined with its long replication time of 12-15 h, none of these theoretical minimal genomes have been tested, even with modern techniques. The publication of the M. genitalium whole-cell model provided the first opportunity to test them, simulating the genome edits in silico. We simulated minimal gene sets from the literature, finding that they produced in silico cells that did not divide. Using knowledge from previous research, we reintroduced specific essential and low essential genes in silico; enabling cellular division. This reinforces the need to identify species-specific low essential genes and their interactions. Any genome designs created using the currently incomplete and fragmented gene essentiality information will very likely require in vivo reintroductions to correct issues and produce dividing cells.

Understanding Metabolic Flux Behaviour in Whole-Cell Model Output.

Whole-cell modelling is a newly expanding field that has many applications in lab experiment design and predictive drug testing. Although whole-cell model output contains a wealth of information, it is complex and high dimensional and thus hard to interpret. Here, we present an analysis pipeline that combines machine learning, dimensionality reduction, and network analysis to interpret and visualise metabolic reaction fluxes from a set of single gene knockouts simulated in the Mycoplasma genitalium whole-cell model. We found that the reaction behaviours show trends that correlate with phenotypic classes of the simulation output, highlighting particular cellular subsystems that malfunction after gene knockouts. From a graphical representation of the metabolic network, we saw that there is a set of reactions that can be used as markers of a phenotypic class, showing their importance within the network. Our analysis pipeline can support the understanding of the complexity of in silico cells without detailed knowledge of the constituent parts, which can help to understand the effects of gene knockouts and, as whole-cell models become more widely built and used, aid genome design.

Minimal Genome Design Algorithms Using Whole-Cell Models.

Synthetic biologists engineer cells and cellular functions using design-build-test cycles; when the task is to extensively engineer entire genomes, the lack of appropriate design tools and biological knowledge about each gene in a cell can lengthen the process, requiring time-consuming and expensive experimental iterations.Whole-cell models represent a new avenue for genome design; the bacteria Mycoplasma genitalium has the first (and currently only published) whole-cell model which combines 28 cellular submodels and represents the integrated functions of every gene and molecule in a cell.We created two minimal genome design algorithms, GAMA and Minesweeper, that produced 1000s of in silico minimal genomes by running simulations on multiple supercomputers. Here we describe the steps to produce in silico cells with reduced genomes, combining minimisation algorithms with whole-cell model simulations.We foresee that the combination of similar algorithms and whole-cell models could later be used for a broad spectrum of genome design applications across cellular species when appropriate models become available.

Harnessing the central dogma for stringent multi-level control of gene expression.

Strictly controlled inducible gene expression is crucial when engineering biological systems where even tiny amounts of a protein have a large impact on function or host cell viability. In these cases, leaky protein production must be avoided, but without affecting the achievable range of expression. Here, we demonstrate how the central dogma offers a simple solution to this challenge. By simultaneously regulating transcription and translation, we show how basal expression of an inducible system can be reduced, with little impact on the maximum expression rate. Using this approach, we create several stringent expression systems displaying >1000-fold change in their output after induction and show how multi-level regulation can suppress transcriptional noise and create digital-like switches between 'on' and 'off' states. These tools will aid those working with toxic genes or requiring precise regulation and propagation of cellular signals, plus illustrate the value of more diverse regulatory designs for synthetic biology.

A Centrifuge-Based Method for Identifying Novel Genetic Traits That Affect Root-Substrate Adhesion in Arabidopsis thaliana.

The physical presence of roots and the compounds they release affect the cohesion between roots and their environment. However, the plant traits that are important for these interactions are unknown and most methods that quantify the contributions of these traits are time-intensive and require specialist equipment and complex substrates. Our lab developed an inexpensive, high-throughput phenotyping assay that quantifies root-substrate adhesion in Arabidopsis thaliana. We now report that this method has high sensitivity and versatility for identifying different types of traits affecting root-substrate adhesion including root hair morphology, vesicle trafficking pathways, and root exudate composition. We describe a practical protocol for conducting this assay and introduce its use in a forward genetic screen to identify novel genes affecting root-substrate interactions. This assay is a powerful tool for identifying and quantifying genetic contributions to cohesion between roots and their environment.

ChipSeg: An Automatic Tool to Segment Bacterial and Mammalian Cells Cultured in Microfluidic Devices.

Extracting quantitative measurements from time-lapse images is necessary in external feedback control applications, where segmentation results are used to inform control algorithms. We describe ChipSeg, a computational tool that segments bacterial and mammalian cells cultured in microfluidic devices and imaged by time-lapse microscopy, which can be used also in the context of external feedback control. The method is based on thresholding and uses the same core functions for both cell types. It allows us to segment individual cells in high cell density microfluidic devices, to quantify fluorescent protein expression over a time-lapse experiment, and to track individual mammalian cells. ChipSeg enables robust segmentation in external feedback control experiments and can be easily customized for other experimental settings and research aims.

Cheetah: A Computational Toolkit for Cybergenetic Control.

Advances in microscopy, microfluidics, and optogenetics enable single-cell monitoring and environmental regulation and offer the means to control cellular phenotypes. The development of such systems is challenging and often results in bespoke setups that hinder reproducibility. To address this, we introduce Cheetah, a flexible computational toolkit that simplifies the integration of real-time microscopy analysis with algorithms for cellular control. Central to the platform is an image segmentation system based on the versatile U-Net convolutional neural network. This is supplemented with functionality to robustly count, characterize, and control cells over time. We demonstrate Cheetah's core capabilities by analyzing long-term bacterial and mammalian cell growth and by dynamically controlling protein expression in mammalian cells. In all cases, Cheetah's segmentation accuracy exceeds that of a commonly used thresholding-based method, allowing for more accurate control signals to be generated. Availability of this easy-to-use platform will make control engineering techniques more accessible and offer new ways to probe and manipulate living cells.

OXI1 kinase is necessary for oxidative burst-mediated signalling in Arabidopsis.

Active oxygen species (AOS) generated in response to stimuli and during development can function as signalling molecules in eukaryotes, leading to specific downstream responses. In plants these include such diverse processes as coping with stress (for example pathogen attack, wounding and oxygen deprivation), abscisic-acid-induced guard-cell closure, and cellular development (for example root hair growth). Despite the importance of signalling via AOS in eukaryotes, little is known about the protein components operating downstream of AOS that mediate any of these processes. Here we show that expression of an Arabidopsis thaliana gene (OXI1) encoding a serine/threonine kinase is induced in response to a wide range of H2O2-generating stimuli. OXI1 kinase activity is itself also induced by H2O2 in vivo. OXI1 is required for full activation of the mitogen-activated protein kinases (MAPKs) MPK3 and MPK6 after treatment with AOS or elicitor and is necessary for at least two very different AOS-mediated processes: basal resistance to Peronospora parasitica infection, and root hair growth. Thus, OXI1 is an essential part of the signal transduction pathway linking oxidative burst signals to diverse downstream responses.

Multiple roles for protein palmitoylation in plants.

Palmitoylation, more correctly known as S-acylation, aids in the regulation of cellular functions including stress response, disease resistance, hormone signalling, cell polarisation, cell expansion and cytoskeletal organization. S-acylation is the reversible addition of fatty acids to proteins, which increases their membrane affinity. Membrane-protein interactions are important for signalling complex formation and signal propagation, protein sequestration and segregation, protein stability, and maintaining polarity within the cell. S-acylation is a dynamic modification that modulates the activity and membrane association of many signalling molecules, including ROP GTPases, heterotrimeric G-proteins and calcium-sensing kinases. Recent advances in methods to study S-acylation are permitting an in-depth examination of its function in plants.

Author Correction: Designing minimal genomes using whole-cell models.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Designing minimal genomes using whole-cell models.

In the future, entire genomes tailored to specific functions and environments could be designed using computational tools. However, computational tools for genome design are currently scarce. Here we present algorithms that enable the use of design-simulate-test cycles for genome design, using genome minimisation as a proof-of-concept. Minimal genomes are ideal for this purpose as they have a simple functional assay whether the cell replicates or not. We used the first (and currently only published) whole-cell model for the bacterium Mycoplasma genitalium. Our computational design-simulate-test cycles discovered novel in silico minimal genomes which, if biologically correct, predict in vivo genomes smaller than JCVI-Syn3.0; a bacterium with, currently, the smallest genome that can be grown in pure culture. In the process, we identified 10 low essential genes and produced evidence for at least two Mycoplasma genitalium in silico minimal genomes. This work brings combined computational and laboratory genome engineering a step closer.

In Vivo Feedback Control of an Antithetic Molecular-Titration Motif in Escherichia coli Using Microfluidics.

We study both in silico and in vivo the real-time feedback control of a molecular titration motif that has been earmarked as a fundamental component of antithetic and multicellular feedback control schemes in E. coli. We show that an external feedback control strategy can successfully regulate the average fluorescence output of a bacterial cell population to a desired constant level in real-time. We also provide in silico evidence that the same strategy can be used to track a time-varying reference signal where the set-point is switched to a different value halfway through the experiment. We use the experimental data to refine and parametrize an in silico model of the motif that can be used as an error computation module in future embedded or multicellular control experiments.

Computer-Aided Whole-Cell Design: Taking a Holistic Approach by Integrating Synthetic With Systems Biology.

Computer-aided design (CAD) for synthetic biology promises to accelerate the rational and robust engineering of biological systems. It requires both detailed and quantitative mathematical and experimental models of the processes to (re)design biology, and software and tools for genetic engineering and DNA assembly. Ultimately, the increased precision in the design phase will have a dramatic impact on the production of designer cells and organisms with bespoke functions and increased modularity. CAD strategies require quantitative models of cells that can capture multiscale processes and link genotypes to phenotypes. Here, we present a perspective on how whole-cell, multiscale models could transform design-build-test-learn cycles in synthetic biology. We show how these models could significantly aid in the design and learn phases while reducing experimental testing by presenting case studies spanning from genome minimization to cell-free systems. We also discuss several challenges for the realization of our vision. The possibility to describe and build whole-cells in silico offers an opportunity to develop increasingly automatized, precise and accessible CAD tools and strategies.

Micro-scale interactions between Arabidopsis root hairs and soil particles influence soil erosion.

Soil is essential for sustaining life on land. Plant roots play a crucial role in stabilising soil and minimising erosion, although these mechanisms are still not completely understood. Consequently, identifying and breeding for plant traits to enhance erosion resistance is challenging. Root hair mutants in Arabidopsis thaliana were studied using three different quantitative methods to isolate their effect on root-soil cohesion. We present compelling evidence that micro-scale interactions of root hairs with surrounding soil increase soil cohesion and reduce erosion. Arabidopsis seedlings with root hairs were more difficult to detach from soil, compost and sterile gel media than those with hairless roots, and it was 10-times harder to erode soil from roots with than without hairs. We also developed a model that can consistently predict the impact root hairs make to soil erosion resistance. Our study thus provides new insight into the mechanisms by which roots maintain soil stability.