Integrating a tailored recurrent neural network with Bayesian experimental design to optimize microbial community functions.

Microbiomes interact dynamically with their environment to perform exploitable functions such as production of valuable metabolites and degradation of toxic metabolites for a wide range of applications in human health, agriculture, and environmental cleanup. Developing computational models to predict the key bacterial species and environmental factors to build and optimize such functions are crucial to accelerate microbial community engineering. However, there is an unknown web of interactions that determine the highly complex and dynamic behavior of these systems, which precludes the development of models based on known mechanisms. By contrast, entirely data-driven machine learning models can produce physically unrealistic predictions and often require significant amounts of experimental data to learn system behavior. We develop a physically-constrained recurrent neural network that preserves model flexibility but is constrained to produce physically consistent predictions and show that it can outperform existing machine learning methods in the prediction of certain experimentally measured species abundance and metabolite concentrations. Further, we present a closed-loop, Bayesian experimental design algorithm to guide data collection by selecting experimental conditions that simultaneously maximize information gain and target microbial community functions. Using a bioreactor case study, we demonstrate how the proposed framework can be used to efficiently navigate a large design space to identify optimal operating conditions. The proposed methodology offers a flexible machine learning approach specifically tailored to optimize microbiome target functions through the sequential design of informative experiments that seek to explore and exploit community functions.

Kinetic Modeling and Parameter Estimation of a Prebiotic Peptide Reaction Network.

Although our understanding of how life emerged on Earth from simple organic precursors is speculative, early precursors likely included amino acids. The polymerization of amino acids into peptides and interactions between peptides are of interest because peptides and proteins participate in complex interaction networks in extant biology. However, peptide reaction networks can be challenging to study because of the potential for multiple species and systems-level interactions between species. We developed and employed a computational network model to describe reactions between amino acids to form di-, tri-, and tetra-peptides. Our experiments were initiated with two of the simplest amino acids, glycine and alanine, mediated by trimetaphosphate-activation and drying to promote peptide bond formation. The parameter estimates for bond formation and hydrolysis reactions in the system were found to be poorly constrained due to a network property known as sloppiness. In a sloppy model, the behavior mostly depends on only a subset of parameter combinations, but there is no straightforward way to determine which parameters should be included or excluded. Despite our inability to determine the exact values of specific kinetic parameters, we could make reasonably accurate predictions of model behavior. In short, our modeling has highlighted challenges and opportunities toward understanding the behaviors of complex prebiotic chemical experiments.

A graph-based modeling framework for tracing hydrological pollutant transport in surface waters.

Anthropogenic pollution of hydrological systems affects diverse communities and ecosystems around the world. Data analytics and modeling tools play a key role in fighting this challenge, as they can help identify key sources as well as trace transport and quantify impact within complex hydrological systems. Several tools exist for simulating and tracing pollutant transport throughout surface waters using detailed physical models; these tools are powerful, but can be computationally intensive, require significant amounts of data to be developed, and require expert knowledge for their use (ultimately limiting application scope). In this work, we present a graph modeling framework - which we call HydroGraphs - for understanding pollutant transport and fate across waterbodies, rivers, and watersheds. This framework uses a simplified representation of hydrological systems that can be constructed based purely on open-source data (National Hydrography Dataset and Watershed Boundary Dataset). The graph representation provides a flexible intuitive approach for capturing connectivity and for identifying upstream pollutant sources and for tracing downstream impacts within small and large hydrological systems. Moreover, the graph representation can facilitate the use of advanced algorithms and tools of graph theory, topology, optimization, and machine learning to aid data analytics and decision-making. We demonstrate the capabilities of our framework by using case studies in the State of Wisconsin; here, we aim to identify upstream nutrient pollutant sources that arise from agricultural practices and trace downstream impacts to waterbodies, rivers, and streams. Our tool ultimately seeks to help stakeholders design effective pollution prevention/mitigation practices and evaluate how surface waters respond to such practices.

Ordering Transitions of Liquid Crystals Triggered by Metal Oxide-catalyzed Reactions of Sulfur Oxide Species.

Liquid crystals (LCs), when supported on reactive surfaces, undergo changes in ordering that can propagate over distances of micrometers, thus providing a general and facile mechanism to amplify atomic-scale transformations on surfaces into the optical scale. While reactions on organic and metal substrates have been coupled to LC-ordering transitions, metal oxide substrates, which offer unique catalytic activities for reactions involving atmospherically important chemical species such as oxidized sulfur species, have not been explored. Here, we investigate this opportunity by designing LCs that contain 4'-cyanobiphenyl-4-carboxylic acid (CBCA) and respond to surface reactions triggered by parts-per-billion concentrations of SO2 gas on anatase (101) substrates. We used electronic structure calculations to predict that the carboxylic acid group of CBCA binds strongly to anatase (101) in a perpendicular orientation, a prediction that we validated in experiments in which CBCA (0.005 mol %) was doped into an LC (4'-n-pentyl-4-biphenylcarbonitrile). Both experiment and computational modeling further demonstrated that SO3-like species, produced by a surface-catalyzed reaction of SO2 with H2O on anatase (101), displace CBCA from the anatase surface, resulting in an orientational transition of the LC. Experiments also reveal the LC response to be highly selective to SO2 over other atmospheric chemical species (including H2O, NH3, H2S, and NO2), in agreement with our computational predictions for anatase (101) surfaces. Overall, we establish that the catalytic activities of metal oxide surfaces offer the basis of a new class of substrates that trigger LCs to undergo ordering transitions in response to chemical species of relevance to atmospheric chemistry.

ADAM: A web platform for graph-based modeling and optimization of supply chains.

Modeling and optimization are essential tasks that arise in the analysis and design of supply chains (SCs). SC models are essential for understanding emergent behavior such as transactions between participants, inherent value of products exchanged, as well as impact of externalities (e.g., policy and climate) and of constraints. Unfortunately, most users of SC models have limited expertise in mathematical optimization, and this hinders the adoption of advanced decision-making tools. In this work, we present ADAM, a web platform that enables the modeling and optimization of SCs. ADAM facilitates modeling by leveraging intuitive and compact graph-based abstractions that allow the user to express dependencies between locations, products, and participants. ADAM model objects serve as repositories of experimental, technology, and socio-economic data; moreover, the graph abstractions facilitate the organization and exchange of models and provides a natural framework for education and outreach. Here, we discuss the graph abstractions and software design principles behind ADAM, its key functional features and workflows, and application examples.

Analysis of incentive policies for phosphorus recovery at livestock facilities in the Great Lakes area.

Livestock operations have been highly intensified over the last decades, resulting in the advent of large concentrated animal feeding operations (CAFOs). Intensification decreases production costs but also leads to substantial environmental impacts. Specifically, nutrient runoff from livestock waste results in eutrophication, harmful algal blooms, and hypoxia. The implementation of nutrient recovery systems in CAFOs can abate nutrient releases and negative ecosystem responses, although they might negatively affect the economic performance of CAFOs. We design and analyze potential incentive policies for the deployment of phosphorus recovery technologies at CAFOs considering the geospatial vulnerability to nutrient pollution. The case study demonstration consists of 2217 CAFOs in the U.S. Great Lakes area. The results reveal that phosphorus recovery is more economically viable in the largest CAFOs due to economies of scale, although they also represent the largest eutrophication threats. For small and medium-scale CAFOs, phosphorus credits progressively improve the profitability of nutrient management systems. The integration of biogas production does not improve the economic performance of phosphorus recovery systems at most of CAFOs, as they lack enough size to be cost-effective. Phosphorus recovery proves to be economically beneficial by comparing the net costs of nutrient management systems with the negative economic impact derived from phosphorus releases. The incentives necessary for avoiding up to 20.7×103 ton/year phosphorus releases and achieve economic neutrality in the Great Lakes area are estimated at $223 million/year. Additionally, the fair distribution of limited incentives is studied using a Nash allocation scheme, determining the break-even point for allocating monetary resources.

Using machine learning and liquid crystal droplets to identify and quantify endotoxins from different bacterial species.

Detection and quantification of bacterial endotoxins is important in a range of health-related contexts, including during pharmaceutical manufacturing of therapeutic proteins and vaccines. Here we combine experimental measurements based on nematic liquid crystalline droplets and machine learning methods to show that it is possible to classify bacterial sources (Escherichia coli, Pseudomonas aeruginosa, Salmonella minnesota) and quantify concentration of endotoxin derived from all three bacterial species present in aqueous solution. The approach uses flow cytometry to quantify, in a high-throughput manner, changes in the internal ordering of micrometer-sized droplets of nematic 4-cyano-4'-pentylbiphenyl triggered by the endotoxins. The changes in internal ordering alter the intensities of light side-scattered (SSC, large-angle) and forward-scattered (FSC, small-angle) by the liquid crystal droplets. A convolutional neural network (Endonet) is trained using the large data sets generated by flow cytometry and shown to predict endotoxin source and concentration directly from the FSC/SSC scatter plots. By using saliency maps, we reveal how EndoNet captures subtle differences in scatter fields to enable classification of bacterial source and quantification of endotoxin concentration over a range that spans eight orders of magnitude (0.01 pg mL-1 to 1 µg mL-1). We attribute changes in scatter fields with bacterial origin of endotoxin, as detected by EndoNet, to the distinct molecular structures of the lipid A domains of the endotoxins derived from the three bacteria. Overall, we conclude that the combination of liquid crystal droplets and EndoNet provides the basis of a promising analytical approach for endotoxins that does not require use of complex biologically-derived reagents (e.g., Limulus amoebocyte lysate).

Valuing economic impact reductions of nutrient pollution from livestock waste.

Nutrient pollution from livestock waste impacts both fresh and marine coastal waters. Harmful algae blooms (HABs) are a common ecosystem-level response to such pollution that is detrimental to both aquatic life and human health and that generates economic losses (e.g., property values and lost tourism). Waste treatment and management technologies are not well established practices due, in part, to the difficulty to attribute economic value to associated social and environmental impacts of nutrient pollution. In this work, we propose a computational framework to quantify the economic impacts of HABs. We demonstrate the advantage of quantifying these impacts through a case study on livestock waste management in the Upper Yahara watershed region (in the state of Wisconsin, USA). Our analysis reveals that every excess kilogram of phosphorus runoff from livestock waste results in total economic losses of 74.5 USD. Furthermore, we use a coordinated market analysis to demonstrate that this economic impact provides a strong enough incentive to activate a nutrient management and valorization market that can help balance phosphorus within the study area. The proposed framework can help state, tribes, and federal regulatory agencies develop regulatory and non-regulatory policies to mitigate the impacts of nutrient pollution.

Scalable nonlinear programming framework for parameter estimation in dynamic biological system models.

We present a nonlinear programming (NLP) framework for the scalable solution of parameter estimation problems that arise in dynamic modeling of biological systems. Such problems are computationally challenging because they often involve highly nonlinear and stiff differential equations as well as many experimental data sets and parameters. The proposed framework uses cutting-edge modeling and solution tools which are computationally efficient, robust, and easy-to-use. Specifically, our framework uses a time discretization approach that: i) avoids repetitive simulations of the dynamic model, ii) enables fully algebraic model implementations and computation of derivatives, and iii) enables the use of computationally efficient nonlinear interior point solvers that exploit sparse and structured linear algebra techniques. We demonstrate these capabilities by solving estimation problems for synthetic human gut microbiome community models. We show that an instance with 156 parameters, 144 differential equations, and 1,704 experimental data points can be solved in less than 3 minutes using our proposed framework (while an off-the-shelf simulation-based solution framework requires over 7 hours). We also create large instances to show that the proposed framework is scalable and can solve problems with up to 2,352 parameters, 2,304 differential equations, and 20,352 data points in less than 15 minutes. The proposed framework is flexible and easy-to-use, can be broadly applied to dynamic models of biological systems, and enables the implementation of sophisticated estimation techniques to quantify parameter uncertainty, to diagnose observability/uniqueness issues, to perform model selection, and to handle outliers.

Coordinated Management of Organic Waste and Derived Products.

We propose a coordination framework for managing urban and rural organic waste in a scalable manner by orchestrating waste exchange, transportation, and transformation into value-added products. The framework is inspired by coordinated management systems that are currently used to operate power grids across the world and that have been instrumental in achieving high levels of efficiency and technological innovation. In the proposed framework, suppliers and consumers of waste and derived products as well as transportation and technology providers bid into a coordination system that is operated by an independent system operator. Allocations and prices for waste and derived products are obtained by the operator by solving a dispatch problem that maximizes the social welfare and that balances supply and demand across a given geographical region. Coordination enables handling of complex constraints and interdependencies that arise from transportation and bio-physico-chemical transformations of waste into products. We prove that the coordination system delivers prices and product allocations that satisfy economic and efficiency properties of a competitive market. The framework is scalable in that it can provide open access that fosters transactions between small and large players in urban and rural areas and over wide geographical regions. Moreover, the framework provides a systematic approach to enable coordinated responses to externalities such as droughts and extreme weather events, to monetize environmental impacts and remediation, to achieve complex social goals such as geographical nutrient balancing, and to justify technology investment and development efforts. Furthermore, the framework can facilitate coordination with electrical, natural gas, water, and transportation, and food distribution infrastructures.

Valuation of Water and Emissions in Energy Systems.

Pricing, incentives, and economic penalties (monetization) are common approaches to control and improve water usage and total direct greenhouse gas emissions (externalities) of energy and other industrial systems. We discuss that homogenous pricing for externalities provides limited flexibility for controlling and improving environmental impacts as different systems are affected differently by externalities. We use trade-off analysis and scalarization techniques to determine marginal prices for water and carbon by taking into account the actual physical and technical limits, stakeholders, and real-time conditions of the system at hand. We demonstrate that high water and emission prices might be needed to control and improve the current system of fixed price for externalities. In addition, a combined heat and power (CHP) system providing hot water and electricity to a real residential building complex is undertaken as case study to demonstrate and describe these concepts. For this CHP system, we found that carbon prices should be increased by a factor of 14 and water prices by a factor of 217 to achieve an optimal compromise between cost, water use, and emissions. Our results point towards the need to consider alternative pricing schemes such as resource bidding (as is done with electricity) that better capture technology trade-offs and push systems towards their efficiency limits.

Spatially explicit methodology for coordinated manure management in shared watersheds.

Increased clustering and consolidation of livestock production systems has been linked to adverse impacts on water quality. This study presents a methodology to optimize manure management within a hydrologic region to minimize agricultural phosphorus (P) loss associated with winter manure application. Spatial and non-spatial data representing livestock, crop, soil, terrain and hydrography were compiled to determine manure P production rates, crop P uptake, existing manure storage capabilities, and transportation distances. Field slope, hydrologic soil group (HSG), and proximity to waterbodies were used to classify crop fields according to their runoff risk for winter-applied manure. We use these data to construct a comprehensive optimization model that identifies optimal location, size, and transportation strategy to achieve environmental and economic goals. The environmental goal was the minimization of daily hauling of manure to environmentally sensitive crop fields, i.e., those classified as high P-loss fields, whereas the economic goal was the minimization of the transportation costs across the entire study area. A case study encompassing two contiguous 10-digit hydrologic unit subwatersheds (HUC-10) in South Central Wisconsin, USA was developed to demonstrate the proposed methodology. Additionally, scenarios representing different management decisions (storage facility maximum volume, and project capital) and production conditions (increased milk production and 20-year future projection) were analyzed to determine their impact on optimal decisions.

The Waste-to-Biomethane Logistic Problem: A Mathematical Optimization Approach.

In this paper, we propose a new mathematical optimization approach to make decisions on the optimal design of the complex logistic system required to produce biogas from waste. We provide a novel and flexible decision-aid tool that allows decision makers to optimally determine the locations of different types of plants (pretreatment, anaerobic digestion, and biomethane liquefaction plants) and pipelines involved in the logistic process, according to a given budget, as well as the most efficient distribution of the products (from waste to biomethane) along the supply chain. The method is based on a mathematical optimization model that we further analyze and that, after reducing the number of variables and constraints without affecting the solutions, is able to solve real-size instances in reasonable CPU times. The proposed methodology is designed to be versatile and adaptable to different situations that arise in the transformation of waste to biogas. The results of our computational experiments, both in synthetic and in a case study instance, prove the validity of our proposal in practical applications. Synthetic instances with up to 200 farms and potential locations for pretreatment plants and 100 potential locations for anaerobic digestion and biomethane liquefaction plants were solved, exactly, within <20 min, whereas the larger instances with 500 farms were solved within <2 h. The CPU times required to solve the real-world instance range from 2 min to 6 h, being highly affected by the given budget to install the plants and the percent of biomethane that is required to be injected in the existing gas network.

Dynamic Multiplexed Control and Modeling of Optogenetic Systems Using the High-Throughput Optogenetic Platform, Lustro.

The ability to control cellular processes using optogenetics is inducer-limited, with most optogenetic systems responding to blue light. To address this limitation, we leverage an integrated framework combining Lustro, a powerful high-throughput optogenetics platform, and machine learning tools to enable multiplexed control over blue light-sensitive optogenetic systems. Specifically, we identify light induction conditions for sequential activation as well as preferential activation and switching between pairs of light-sensitive split transcription factors in the budding yeast, Saccharomyces cerevisiae. We use the high-throughput data generated from Lustro to build a Bayesian optimization framework that incorporates data-driven learning, uncertainty quantification, and experimental design to enable the prediction of system behavior and the identification of optimal conditions for multiplexed control. This work lays the foundation for designing more advanced synthetic biological circuits incorporating optogenetics, where multiple circuit components can be controlled using designer light induction programs, with broad implications for biotechnology and bioengineering.

Multi-site, multi-pollutant atmospheric data analysis using Riemannian geometry.

We demonstrate the benefits of using Riemannian geometry in the analysis of multi-site, multi-pollutant atmospheric monitoring data. Our approach uses covariance matrices to encode spatio-temporal variability and correlations of multiple pollutants at different sites and times. A key property of covariance matrices is that they lie on a Riemannian manifold and one can exploit this property to facilitate dimensionality reduction, outlier detection, and spatial interpolation. Specifically, the transformation of data using Reimannian geometry provides a better data surface for interpolation and assessment of outliers compared to traditional data analysis tools that assume Euclidean geometry. We demonstrate the utility of using Riemannian geometry by analyzing a full year of atmospheric monitoring data collected from 34 monitoring stations in Beijing, China.

Dynamic Multiplexed Control and Modeling of Optogenetic Systems Using the High-Throughput Optogenetic Platform, Lustro.

The ability to control cellular processes using optogenetics is inducer-limited, with most optogenetic systems responding to blue light. To address this limitation we leverage an integrated framework combining Lustro, a powerful high-throughput optogenetics platform, and machine learning tools to enable multiplexed control over blue light-sensitive optogenetic systems. Specifically, we identify light induction conditions for sequential activation as well as preferential activation and switching between pairs of light-sensitive spit transcription factors in the budding yeast, Saccharomyces cerevisiae . We use the high-throughput data generated from Lustro to build a Bayesian optimization framework that incorporates data-driven learning, uncertainty quantification, and experimental design to enable the prediction of system behavior and the identification of optimal conditions for multiplexed control. This work lays the foundation for designing more advanced synthetic biological circuits incorporating optogenetics, where multiple circuit components can be controlled using designer light induction programs, with broad implications for biotechnology and bioengineering.

Topological Analysis of Molecular Dynamics Simulations using the Euler Characteristic.

Molecular dynamics (MD) simulations are used in diverse scientific and engineering fields such as drug discovery, materials design, separations, biological systems, and reaction engineering. These simulations generate highly complex data sets that capture the 3D spatial positions, dynamics, and interactions of thousands of molecules. Analyzing MD data sets is key for understanding and predicting emergent phenomena and in identifying key drivers and tuning design knobs of such phenomena. In this work, we show that the Euler characteristic (EC) provides an effective topological descriptor that facilitates MD analysis. The EC is a versatile, low-dimensional, and easy-to-interpret descriptor that can be used to reduce, analyze, and quantify complex data objects that are represented as graphs/networks, manifolds/functions, and point clouds. Specifically, we show that the EC is an informative descriptor that can be used for machine learning and data analysis tasks such as classification, visualization, and regression. We demonstrate the benefits of the proposed approach through case studies that aim to understand and predict the hydrophobicity of self-assembled monolayers and the reactivity of complex solvent environments.

Hydroformylation of pyrolysis oils to aldehydes and alcohols from polyolefin waste.

Waste plastics are an abundant feedstock for the production of renewable chemicals. Pyrolysis of waste plastics produces pyrolysis oils with high concentrations of olefins (>50 weight %). The traditional petrochemical industry uses several energy-intensive steps to produce olefins from fossil feedstocks such as naphtha, natural gas, and crude oil. In this work, we demonstrate that pyrolysis oil can be used to produce aldehydes through hydroformylation, taking advantage of the olefin functionality. These aldehydes can then be reduced to mono- and dialcohols, oxidized to mono- and dicarboxylic acids, or aminated to mono- and diamines by using homogeneous and heterogeneous catalysis. This route produces high-value oxygenated chemicals from low-value postconsumer recycled polyethylene. We project that the chemicals produced by this route could lower greenhouse gas emissions ~60% compared with their production through petroleum feedstocks.

Techno-Economic and life cycle assessment of standalone Single-Stream material recovery facilities in the United states.

Material Recovery Facilities (MRFs) are crucial players in achieving a circular economy. MRFs receive complex waste streams and separate valuable recyclables from these mixtures. This study conducts techno-economic analysis (TEA) to estimate the net present value (NPV) and life cycle assessment (LCA) to estimate different environmental impacts of a commercial scale standalone, single-stream MRF to assess the economic feasibility and environmental impacts of recovering valuable recyclables from an MRF processing 120,000 tonnes per year (t/y). The TEA employs a discounted cash flow rate of return (DCFROR) analysis over a 20-year facility lifetime, along with a sensitivity analysis on the impact of different operating and economic parameters. Results show that the total fixed cost of building the MRF facility is $23 MM, and the operating cost is $45.48/tonne. The NPV of the MRF can vary from $3.57 MM to $60 MM, while 100-year global warming potential can range from 5.98 to 8.53 kg carbon dioxide equivalents (CO2-eq) per tonne of MSW. We have also found that MSW composition (arising from regional effects) significantly impacts costs, 100-year global warming potential, and other impact categories such as acidification potential, eutrophication potential, ecotoxicity, ozone depletion, photochemical oxidation, carcinogenic effects, and non-carcinogenic effects. Sensitivity and uncertainty analysis indicate that waste composition and market prices significantly impact the profitability of the MRF, and the waste composition mostly impacts global warming potential. Our analysis also indicates that facility capacity, fixed capital cost, and waste tipping fees are vital parameters that affect the economic viability of MRF operations.

Decoding Optical Responses of Contact-Printed Arrays of Thermotropic Liquid Crystals Using Machine Learning: Detection and Reporting of Aqueous Amphiphiles with Enhanced Sensitivity and Selectivity.

Surfactants and other amphiphilic molecules are used extensively in household products, industrial processes, and biological applications and are also common environmental contaminants; as such, methods that can detect, sense, or quantify them are of great practical relevance. Aqueous emulsions of thermotropic liquid crystals (LCs) can exhibit distinctive optical responses in the presence of surfactants and have thus emerged as sensitive, rapid, and inexpensive sensors or reporters of environmental amphiphiles. However, many existing LC-in-water emulsions require the use of complicated or expensive instrumentation for quantitative characterization owing to variations in optical responses among individual LC droplets. In many cases, the responses of LC droplets are also analyzed by human inspection, which can miss subtle color or topological changes encoded in LC birefringence patterns. Here, we report an LC-based surfactant sensing platform that takes a step toward addressing several of these issues and can reliably predict concentrations and types of surfactants in aqueous solutions. Our approach uses surface-immobilized, microcontact-printed arrays of micrometer-scale droplets of thermotropic LCs and hierarchical convolutional neural networks (CNNs) to automatically extract and decode rich information about topological defects and color patterns available in optical micrographs of LC droplets to classify and quantify adsorbed surfactants. In addition, we report computational capabilities to determine relevant optical features extracted by the CNN from LC micrographs, which can provide insights into surfactant adsorption phenomena at LC-water interfaces. Overall, the combination of microcontact-printed LC arrays and machine learning provides a convenient and robust platform that could prove useful for developing high-throughput sensors for on-site testing of environmentally or biologically relevant amphiphiles.