Deterministic evolution and stringent selection during preneoplasia.

The earliest events during human tumour initiation, although poorly characterized, may hold clues to malignancy detection and prevention1. Here we model occult preneoplasia by biallelic inactivation of TP53, a common early event in gastric cancer, in human gastric organoids. Causal relationships between this initiating genetic lesion and resulting phenotypes were established using experimental evolution in multiple clonally derived cultures over 2 years. TP53 loss elicited progressive aneuploidy, including copy number alterations and structural variants prevalent in gastric cancers, with evident preferred orders. Longitudinal single-cell sequencing of TP53-deficient gastric organoids similarly indicates progression towards malignant transcriptional programmes. Moreover, high-throughput lineage tracing with expressed cellular barcodes demonstrates reproducible dynamics whereby initially rare subclones with shared transcriptional programmes repeatedly attain clonal dominance. This powerful platform for experimental evolution exposes stringent selection, clonal interference and a marked degree of phenotypic convergence in premalignant epithelial organoids. These data imply predictability in the earliest stages of tumorigenesis and show evolutionary constraints and barriers to malignant transformation, with implications for earlier detection and interception of aggressive, genome-instable tumours.

Phenotypic plasticity and genetic control in colorectal cancer evolution.

Genetic and epigenetic variation, together with transcriptional plasticity, contribute to intratumour heterogeneity1. The interplay of these biological processes and their respective contributions to tumour evolution remain unknown. Here we show that intratumour genetic ancestry only infrequently affects gene expression traits and subclonal evolution in colorectal cancer (CRC). Using spatially resolved paired whole-genome and transcriptome sequencing, we find that the majority of intratumour variation in gene expression is not strongly heritable but rather 'plastic'. Somatic expression quantitative trait loci analysis identified a number of putative genetic controls of expression by cis-acting coding and non-coding mutations, the majority of which were clonal within a tumour, alongside frequent structural alterations. Consistently, computational inference on the spatial patterning of tumour phylogenies finds that a considerable proportion of CRCs did not show evidence of subclonal selection, with only a subset of putative genetic drivers associated with subclone expansions. Spatial intermixing of clones is common, with some tumours growing exponentially and others only at the periphery. Together, our data suggest that most genetic intratumour variation in CRC has no major phenotypic consequence and that transcriptional plasticity is, instead, widespread within a tumour.

Chaos in synthetic microbial communities.

Predictability is a fundamental requirement in biological engineering. As we move to building coordinated multicellular systems, the potential for such systems to display chaotic behaviour becomes a concern. Therefore understanding which systems show chaos is an important design consideration. We developed a methodology to explore the potential for chaotic dynamics in small microbial communities governed by resource competition, intercellular communication and competitive bacteriocin interactions. Our model selection pipeline uses Approximate Bayesian Computation to first identify oscillatory behaviours as a route to finding chaotic behaviour. We have shown that we can expect to find chaotic states in relatively small synthetic microbial systems, understand the governing dynamics and provide insights into how to control such systems. This work is the first to query the existence of chaotic behaviour in synthetic microbial communities and has important ramifications for the fields of biotechnology, bioprocessing and synthetic biology.

Deep reinforcement learning for optimal experimental design in biology.

The field of optimal experimental design uses mathematical techniques to determine experiments that are maximally informative from a given experimental setup. Here we apply a technique from artificial intelligence-reinforcement learning-to the optimal experimental design task of maximizing confidence in estimates of model parameter values. We show that a reinforcement learning approach performs favourably in comparison with a one-step ahead optimisation algorithm and a model predictive controller for the inference of bacterial growth parameters in a simulated chemostat. Further, we demonstrate the ability of reinforcement learning to train over a distribution of parameters, indicating that this approach is robust to parametric uncertainty.

Engineered acetoacetate-inducible whole-cell biosensors based on the AtoSC two-component system.

Whole-cell biosensors hold potential in a variety of industrial, medical, and environmental applications. These biosensors can be constructed through the repurposing of bacterial sensing mechanisms, including the common two-component system (TCS). Here we report on the construction of a range of novel biosensors that are sensitive to acetoacetate, a molecule that plays a number of roles in human health and biology. These biosensors are based on the AtoSC TCS. An ordinary differential equation model to describe the action of the AtoSC TCS was developed and sensitivity analysis of this model used to help inform biosensor design. The final collection of biosensors constructed displayed a range of switching behaviours at physiologically relevant acetoacetate concentrations and can operate in several Escherichia coli host strains. It is envisaged that these biosensor strains will offer an alternative to currently available commercial strip tests and, in future, may be adopted for more complex in vivo or industrial monitoring applications.

Deep reinforcement learning for the control of microbial co-cultures in bioreactors.

Multi-species microbial communities are widespread in natural ecosystems. When employed for biomanufacturing, engineered synthetic communities have shown increased productivity in comparison with monocultures and allow for the reduction of metabolic load by compartmentalising bioprocesses between multiple sub-populations. Despite these benefits, co-cultures are rarely used in practice because control over the constituent species of an assembled community has proven challenging. Here we demonstrate, in silico, the efficacy of an approach from artificial intelligence-reinforcement learning-for the control of co-cultures within continuous bioreactors. We confirm that feedback via a trained reinforcement learning agent can be used to maintain populations at target levels, and that model-free performance with bang-bang control can outperform a traditional proportional integral controller with continuous control, when faced with infrequent sampling. Further, we demonstrate that a satisfactory control policy can be learned in one twenty-four hour experiment by running five bioreactors in parallel. Finally, we show that reinforcement learning can directly optimise the output of a co-culture bioprocess. Overall, reinforcement learning is a promising technique for the control of microbial communities.

Catch my drift? Making sense of genomic intra-tumour heterogeneity.

The cancer genome is shaped by three components of the evolutionary process: mutation, selection and drift. While many studies have focused on the first two components, the role of drift in cancer evolution has received little attention. Drift occurs when all individuals in the population have the same likelihood of producing surviving offspring, and so by definition a drifting population is one that is evolving neutrally. Here we focus on how neutral evolution is manifested in the cancer genome. We discuss how neutral passenger mutations provide a magnifying glass that reveals the evolutionary dynamics underpinning cancer development, and outline how statistical inference can be used to quantify these dynamics from sequencing data. We argue that only after we understand the impact of neutral drift on the genome can we begin to make full sense of clonal selection. This article is part of a Special Issue entitled: Evolutionary principles - heterogeneity in cancer? Edited by Dr. Robert A. Gatenby.

Crypt fusion as a homeostatic mechanism in the human colon.

OBJECTIVE: The crypt population in the human intestine is dynamic: crypts can divide to produce two new daughter crypts through a process termed crypt fission, but whether this is balanced by a second process to remove crypts, as recently shown in mouse models, is uncertain. We examined whether crypt fusion (the process of two neighbouring crypts fusing into a single daughter crypt) occurs in the human colon. DESIGN: We used somatic alterations in the gene cytochrome c oxidase (CCO) as lineage tracing markers to assess the clonality of bifurcating colon crypts (n=309 bifurcating crypts from 13 patients). Mathematical modelling was used to determine whether the existence of crypt fusion can explain the experimental data, and how the process of fusion influences the rate of crypt fission. RESULTS: In 55% (21/38) of bifurcating crypts in which clonality could be assessed, we observed perfect segregation of clonal lineages to the respective crypt arms. Mathematical modelling showed that this frequency of perfect segregation could not be explained by fission alone (p<10-20). With the rates of fission and fusion taken to be approximately equal, we then used the distribution of CCO-deficient patch size to estimate the rate of crypt fission, finding a value of around 0.011 divisions/crypt/year. CONCLUSIONS: We have provided the evidence that human colonic crypts undergo fusion, a potential homeostatic process to regulate total crypt number. The existence of crypt fusion in the human colon adds a new facet to our understanding of the highly dynamic and plastic phenotype of the colonic epithelium.

Computing with biological switches and clocks.

The complex dynamics of biological systems is primarily driven by molecular interactions that underpin the regulatory networks of cells. These networks typically contain positive and negative feedback loops, which are responsible for switch-like and oscillatory dynamics, respectively. Many computing systems rely on switches and clocks as computational modules. While the combination of such modules in biological systems leads to a variety of dynamical behaviours, it is also driving development of new computing algorithms. Here we present a historical perspective on computation by biological systems, with a focus on switches and clocks, and discuss parallels between biology and computing. We also outline our vision for the future of biological computing.

A theoretical framework for the regulation of Shh morphogen-controlled gene expression.

How morphogen gradients govern the pattern of gene expression in developing tissues is not well understood. Here, we describe a statistical thermodynamic model of gene regulation that combines the activity of a morphogen with the transcriptional network it controls. Using Sonic hedgehog (Shh) patterning of the ventral neural tube as an example, we show that the framework can be used together with the principled parameter selection technique of approximate Bayesian computation to obtain a dynamical model that accurately predicts tissue patterning. The analysis indicates that, for each target gene regulated by Gli, which is the transcriptional effector of Shh signalling, there is a neutral point in the gradient, either side of which altering the Gli binding affinity has opposite effects on gene expression. This explains recent counterintuitive experimental observations. The approach is broadly applicable and provides a unifying framework to explain the temporospatial pattern of morphogen-regulated gene expression.

Mechanistic Modelling and Bayesian Inference Elucidates the Variable Dynamics of Double-Strand Break Repair.

DNA double-strand breaks are lesions that form during metabolism, DNA replication and exposure to mutagens. When a double-strand break occurs one of a number of repair mechanisms is recruited, all of which have differing propensities for mutational events. Despite DNA repair being of crucial importance, the relative contribution of these mechanisms and their regulatory interactions remain to be fully elucidated. Understanding these mutational processes will have a profound impact on our knowledge of genomic instability, with implications across health, disease and evolution. Here we present a new method to model the combined activation of non-homologous end joining, single strand annealing and alternative end joining, following exposure to ionising radiation. We use Bayesian statistics to integrate eight biological data sets of double-strand break repair curves under varying genetic knockouts and confirm that our model is predictive by re-simulating and comparing to additional data. Analysis of the model suggests that there are at least three disjoint modes of repair, which we assign as fast, slow and intermediate. Our results show that when multiple data sets are combined, the rate for intermediate repair is variable amongst genetic knockouts. Further analysis suggests that the ratio between slow and intermediate repair depends on the presence or absence of DNA-PKcs and Ku70, which implies that non-homologous end joining and alternative end joining are not independent. Finally, we consider the proportion of double-strand breaks within each mechanism as a time series and predict activity as a function of repair rate. We outline how our insights can be directly tested using imaging and sequencing techniques and conclude that there is evidence of variable dynamics in alternative repair pathways. Our approach is an important step towards providing a unifying theoretical framework for the dynamics of DNA repair processes.

Clinical drug resistance linked to interconvertible phenotypic and functional states of tumor-propagating cells in multiple myeloma.

The phenotype and function of cells enriched in tumor-propagating activity and their relationship to the phenotypic architecture in multiple myeloma (MM) are controversial. Here, in a cohort of 30 patients, we show that MM composes 4 hierarchically organized, clonally related subpopulations, which, although phenotypically distinct, share the same oncogenic chromosomal abnormalities as well as immunoglobulin heavy chain complementarity region 3 area sequence. Assessed in xenograft assays, myeloma-propagating activity is the exclusive property of a population characterized by its ability for bidirectional transition between the dominant CD19(-)CD138(+) plasma cell (PC) and a low frequency CD19(-)CD138(-) subpopulation (termed Pre-PC); in addition, Pre-PCs are more quiescent and unlike PCs, are primarily localized at extramedullary sites. As shown by gene expression profiling, compared with PCs, Pre-PCs are enriched in epigenetic regulators, suggesting that epigenetic plasticity underpins the phenotypic diversification of myeloma-propagating cells. Prospective assessment in paired, pretreatment, and posttreatment bone marrow samples shows that Pre-PCs are up to 300-fold more drug-resistant than PCs. Thus, clinical drug resistance in MM is linked to reversible, bidirectional phenotypic transition of myeloma-propagating cells. These novel biologic insights have important clinical implications in relation to assessment of minimal residual disease and development of alternative therapeutic strategies in MM.

Model selection in systems biology depends on experimental design.

Experimental design attempts to maximise the information available for modelling tasks. An optimal experiment allows the inferred models or parameters to be chosen with the highest expected degree of confidence. If the true system is faithfully reproduced by one of the models, the merit of this approach is clear - we simply wish to identify it and the true parameters with the most certainty. However, in the more realistic situation where all models are incorrect or incomplete, the interpretation of model selection outcomes and the role of experimental design needs to be examined more carefully. Using a novel experimental design and model selection framework for stochastic state-space models, we perform high-throughput in-silico analyses on families of gene regulatory cascade models, to show that the selected model can depend on the experiment performed. We observe that experimental design thus makes confidence a criterion for model choice, but that this does not necessarily correlate with a model's predictive power or correctness. Finally, in the special case of linear ordinary differential equation (ODE) models, we explore how wrong a model has to be before it influences the conclusions of a model selection analysis.

On optimality of kernels for approximate Bayesian computation using sequential Monte Carlo.

Approximate Bayesian computation (ABC) has gained popularity over the past few years for the analysis of complex models arising in population genetics, epidemiology and system biology. Sequential Monte Carlo (SMC) approaches have become work-horses in ABC. Here we discuss how to construct the perturbation kernels that are required in ABC SMC approaches, in order to construct a sequence of distributions that start out from a suitably defined prior and converge towards the unknown posterior. We derive optimality criteria for different kernels, which are based on the Kullback-Leibler divergence between a distribution and the distribution of the perturbed particles. We will show that for many complicated posterior distributions, locally adapted kernels tend to show the best performance. We find that the added moderate cost of adapting kernel functions is easily regained in terms of the higher acceptance rate. We demonstrate the computational efficiency gains in a range of toy examples which illustrate some of the challenges faced in real-world applications of ABC, before turning to two demanding parameter inference problems in molecular biology, which highlight the huge increases in efficiency that can be gained from choice of optimal kernels. We conclude with a general discussion of the rational choice of perturbation kernels in ABC SMC settings.

Bayesian design of synthetic biological systems.

Here we introduce a new design framework for synthetic biology that exploits the advantages of Bayesian model selection. We will argue that the difference between inference and design is that in the former we try to reconstruct the system that has given rise to the data that we observe, whereas in the latter, we seek to construct the system that produces the data that we would like to observe, i.e., the desired behavior. Our approach allows us to exploit methods from Bayesian statistics, including efficient exploration of models spaces and high-dimensional parameter spaces, and the ability to rank models with respect to their ability to generate certain types of data. Bayesian model selection furthermore automatically strikes a balance between complexity and (predictive or explanatory) performance of mathematical models. To deal with the complexities of molecular systems we employ an approximate Bayesian computation scheme which only requires us to simulate from different competing models to arrive at rational criteria for choosing between them. We illustrate the advantages resulting from combining the design and modeling (or in silico prototyping) stages currently seen as separate in synthetic biology by reference to deterministic and stochastic model systems exhibiting adaptive and switch-like behavior, as well as bacterial two-component signaling systems.

Emergent digital bio-computation through spatial diffusion and engineered bacteria.

Biological computing is a promising field with potential applications in biosafety, environmental monitoring, and personalized medicine. Here we present work on the design of bacterial computers using spatial patterning to process information in the form of diffusible morphogen-like signals. We demonstrate, mathematically and experimentally, that single, modular, colonies can perform simple digital logic, and that complex functions can be built by combining multiple colonies, removing the need for further genetic engineering. We extend our experimental system to incorporate sender colonies as morphogen sources, demonstrating how one might integrate different biochemical inputs. Our approach will open up ways to perform biological computation, with applications in bioengineering, biomaterials and biosensing. Ultimately, these computational bacterial communities will help us explore information processing in natural biological systems.

A bacteriocin expression platform for targeting pathogenic bacterial species.

Bacteriocins are antimicrobial peptides that are naturally produced by many bacteria. They hold great potential in the fight against antibiotic resistant bacteria, including ESKAPE pathogens. Engineered live biotherapeutic products (eLBPs) that secrete bacteriocins can be created to deliver targeted bacteriocin production. Here we develop a modular bacteriocin secretion platform that can be used to express and secrete multiple bacteriocins from non-pathogenic Escherichia coli host strains. As a proof of concept we create Enterocin A (EntA) and Enterocin B (EntB) secreting strains that show strong antimicrobial activity against Enterococcus faecalis and Enterococcus faecium in vitro, and characterise this activity in both solid culture and liquid co-culture. We then develop a Lotka-Volterra model that can be used to capture the interactions of these competitor strains. We show that simultaneous exposure to EntA and EntB can delay Enterococcus growth. Our system has the potential to be used as an eLBP to secrete additional bacteriocins for the targeted killing of pathogenic bacteria.

Reduced burden of very large and rare CNVs in bipolar affective disorder.

OBJECTIVES: Large, rare chromosomal copy number variants (CNVs) have been shown to increase the risk for schizophrenia and other neuropsychiatric disorders including autism, attention-deficit hyperactivity disorder, learning difficulties, and epilepsy. Their role in bipolar disorder (BD) is less clear. There are no reports of an increase in large, rare CNVs in BD in general, but some have reported an increase in early-onset cases. We previously found that the rate of such CNVs in individuals with BD was not increased, even in early-onset cases. Our aim here was to examine the rate of large rare CNVs in BD in comparison with a new large independent reference sample from the same country. METHODS: We studied the CNVs in a case-control sample consisting of 1,650 BD cases (reported previously) and 10,259 reference individuals without a known psychiatric disorder who took part in the original Wellcome Trust Case Control Consortium (WTCCC) study. The 10,259 reference individuals were affected with six non-psychiatric disorders (coronary artery disease, types 1 and 2 diabetes, hypertension, Crohn's disease, and rheumatoid arthritis). Affymetrix 500K array genotyping data were used to call the CNVs. RESULTS: The rate of CNVs > 100 kb was not statistically different between cases and controls. The rate of very large (defined as > 1 Mb) and rare (< 1%) CNVs was significantly lower in patients with BD compared with the reference group. CNV loci associated with schizophrenia were not enriched in BD and, in fact, cases of BD had the lowest number of such CNVs compared with any of the WTCCC cohorts; this finding held even for the early-onset BD cases. CONCLUSIONS: Schizophrenia and BD differ with respect to CNV burden and association with specific CNVs. Our findings support the hypothesis that BD is etiologically distinct from schizophrenia with respect to large, rare CNVs and the accompanying associated neurodevelopmental abnormalities.

CNETML: maximum likelihood inference of phylogeny from copy number profiles of multiple samples.

Phylogenetic trees based on copy number profiles from multiple samples of a patient are helpful to understand cancer evolution. Here, we develop a new maximum likelihood method, CNETML, to infer phylogenies from such data. CNETML is the first program to jointly infer the tree topology, node ages, and mutation rates from total copy numbers of longitudinal samples. Our extensive simulations suggest CNETML performs well on copy numbers relative to ploidy and under slight violation of model assumptions. The application of CNETML to real data generates results consistent with previous discoveries and provides novel early copy number events for further investigation.