Search

Multi-stage screening pipelines are ubiquitous throughout experimental and computational science. Much of the effort in developing screening pipelines focuses on improving generative methods or surrogate models in an attempt to make each screening step effective for a specific application. Little focus has been placed on characterizing generic screening pipeline performance with respect to the problem or problem parameters. Here, we develop models and algorithms to codify and simulate features and properties of the screening procedure in general. We outline and model common problem settings and potential opportunities to perform decision-making under uncertainty to optimize the execution of screening pipelines. We then illustrate the models and algorithms through several simulation studies. We finally show how such studies can provide a quantification of the screening pipeline performance with respect to problem parameters, specifically identifying the significance of stage-wise covariance structure. We show how such structure can lead to qualitatively different screening behaviors, and how screening can even perform worse than random in some cases.

Summary Solutions to many of the world's problems depend upon materials research and development. However, advanced materials can take decades to discover and decades more to fully deploy. Humans and robots have begun to partner to advance science and technology orders of magnitude faster than humans do today through the development and exploitation of closed-loop, autonomous experimentation systems. This review discusses the specific challenges and opportunities related to materials discovery and development that will emerge from this new paradigm. Our perspective incorporates input from stakeholders in academia, industry, government laboratories, and funding agencies. We outline the current status, barriers, and needed investments, culminating with a vision for the path forward. We intend the article to spark interest in this emerging research area and to motivate potential practitioners by illustrating early successes. We also aspire to encourage a creative reimagining of the next generation of materials science infrastructure. To this end, we frame future investments in materials science and technology, hardware and software infrastructure, artificial intelligence and autonomy methods, and critical workforce development for autonomous research.

Autonomous robotic experimentation strategies are rapidly rising in use because, without the need for user intervention, they can efficiently and precisely converge onto optimal intrinsic and extrinsic synthesis conditions for a wide range of emerging materials. However, as the material syntheses become more complex, the meta-decisions of artificial intelligence (AI)-guided decision-making algorithms used in autonomous platforms become more important. In this work, a surrogate model is developed using data from over 1000 in-house conducted syntheses of metal halide perovskite quantum dots in a self-driven modular microfluidic material synthesizer. The model is designed to represent the global failure rate, unfeasible regions of the synthesis space, synthesis ground truth, and sampling noise of a real robotic material synthesis system with multiple output parameters (peak emission, emission linewidth, and quantum yield). With this model, over 150 AI-guided decision-making strategies within a single-period horizon reinforcement learning framework are automatically explored across more than 600 000 simulated experiments – the equivalent of 7.5 years of continuous robotic operation and 400 L of reagents – to identify the most effective methods for accelerated materials development with multiple objectives. Specifically, the structure and meta-decisions of an ensemble neural network-based material development strategy are investigated, which offers a favorable technique for intelligently and efficiently navigating a complex material synthesis space with multiple targets. The developed ensemble neural network-based decision-making algorithm enables more efficient material formulation optimization in a no prior information environment than well-established algorithms., A surrogate model is designed to represent a microfluidic material synthesis system using 1000 automatically conducted experiments. With this model, over 600 000 experiments are simulated to optimize an AI-guided material synthesis algorithm.

Machine learning (ML) and artificial intelligence (AI) are quickly becoming commonplace in materials research. In addition to the standard workflow of fitting a model to a large set of data in order to make predictions, the materials community is finding novel and meaningful ways to integrate AI within their work. This has led to an acceleration not only of materials design and discovery, but also of other aspects of materials research as well, including faster computational models, the development of autonomous and intelligent “robot researchers,” and the automatic discovery of physical models. In this issue, we highlight a few of these applications and argue that AI/ML is delivering real-world, practical solutions to materials problems. It is also clear that we need AI/ML methods and models, “dialects” that are better adapted to materials research.

Microbes embedded in hydrogels comprise one form of living material. Discovering formulations that balance potentially competing mechanical and biological properties in living hydrogels—for example gel time of the hydrogel formulation and viability of the embedded organisms—can be challenging. In this work, a pipeline is developed to automate characterization of the gel time of hydrogel formulations. Using this pipeline, living materials comprised of enzymatically crosslinked silk and embedded E. coli—formulated from within a 4D parameter space—are engineered to gel within a pre-selected timeframe. Gelation time is estimated using a novel adaptation of microrheology analysis using differential dynamic microscopy (DDM). In order to expedite the discovery of gelation regime boundaries, Bayesian machine learning models are deployed with optimal decision-making under uncertainty. The rate of learning is observed to vary between AI-assisted planning and human planning, with the fastest rate occurring during AI-assisted planning following a round of human planning. For a subset of formulations gelling within a targeted timeframe of 5-15 minutes, fluorophore production within the embedded cells is substantially similar across treatments, evidencing that gel time can be tuned independent of other material properties—at least over a finite range—while maintaining biological activity.

Summary Autonomous experimentation (AE) accelerates research by combining automation and machine learning to perform experiments intelligently and rapidly in a sequential fashion. While AE systems are most needed to study properties that cannot be predicted analytically or computationally, even imperfect predictions can in principle be useful. Here, we investigate whether imperfect data from simulation can accelerate AE using a case study on the mechanics of additively manufactured structures. Initially, we study resilience, a property that is well-predicted by finite element analysis (FEA), and find that FEA can be used to build a Bayesian prior and experimental data can be integrated using discrepancy modeling to reduce the number of needed experiments ten-fold. Next, we study toughness, a property not well-predicted by FEA and find that FEA can still improve learning by transforming experimental data and guiding experiment selection. These results highlight multiple ways that simulation can improve AE through transfer learning., Graphical abstract, Highlights • Simulation and autonomous experimentation were combined to accelerate research • Resilience, which was accurately simulated, was learned in 10 x fewer experiments • Simulating related properties, i.e. yield force, accelerated learning toughness • Simulation was introduced to an experimental learning loop using transfer learning, Mechanical Property; Computational Method in Materials Science; Simulation in Materials Science

Several two-dimensional (2D) materials such as graphene, molybdenum disulfide, or boron nitride are emerging as alternatives for lubrication additives to control friction and wear at the interface. On the other hand, the initiative to accelerate materials discovery through data-driven computational methods has identified numerous novel topologies and families of 2D materials that can potentially be designed as low-friction additives. Hence, generating a structure–property (friction) correlations for 2D material-based additives that present a large variation in atomic composition is the next big challenge. Herein, we present a machine learning (ML) method using the Bayesian modeling and transfer learning approach to predict the maximum energy barrier (MEB) of the potential surface energy (correlated to intrinsic friction) of ten different 2D materials that were previously unexplored for their tribological properties. The descriptors (or properties) required to train the ML model with high accuracy are identified by taking into account the established physical models for dissipation in 2D materials. As a result, a difference of less than 8% in MEB values as predicted via the ML model presented here and the PES profiles generated using molecular dynamics simulations, for a select few 2D materials, was obtained. The model also enabled the identification of material properties that present the highest sensitivity to the corrugated potential, hence enabling the development of design routes for the synthesis of 2D materials with optimal tribological properties.

Automated testing, Bayesian optimization, and additive manufacturing combine for the autonomous design of structures., While additive manufacturing (AM) has facilitated the production of complex structures, it has also highlighted the immense challenge inherent in identifying the optimum AM structure for a given application. Numerical methods are important tools for optimization, but experiment remains the gold standard for studying nonlinear, but critical, mechanical properties such as toughness. To address the vastness of AM design space and the need for experiment, we develop a Bayesian experimental autonomous researcher (BEAR) that combines Bayesian optimization and high-throughput automated experimentation. In addition to rapidly performing experiments, the BEAR leverages iterative experimentation by selecting experiments based on all available results. Using the BEAR, we explore the toughness of a parametric family of structures and observe an almost 60-fold reduction in the number of experiments needed to identify high-performing structures relative to a grid-based search. These results show the value of machine learning in experimental fields where data are sparse.

Microbes embedded in hydrogels comprise one form of living material. Discovering formulations that balance potentially competing for mechanical and biological properties in living hydrogels-for example, gel time of the hydrogel formulation and viability of the embedded organisms-can be challenging. In this study, a pipeline is developed to automate the characterization of the gel time of hydrogel formulations. Using this pipeline, living materials comprised of enzymatically crosslinked silk and embedded E. coli-formulated from within a 4D parameter space-are engineered to gel within a pre-selected timeframe. Gelation time is estimated using a novel adaptation of microrheology analysis using differential dynamic microscopy (DDM). In order to expedite the discovery of gelation regime boundaries, Bayesian machine learning models are deployed with optimal decision-making under uncertainty. The rate of learning is observed to vary between artificial intelligence (AI)-assisted planning and human planning, with the fastest rate occurring during AI-assisted planning following a round of human planning. For a subset of formulations gelling within a targeted timeframe of 5-15 min, fluorophore production within the embedded cells is substantially similar across treatments, evidencing that gel time can be tuned independent of other material properties-at least over a finite range-while maintaining biological activity.

Autonomous experimentation (AE) accelerates research by combining automation and machine learning to perform experiments intelligently and rapidly in a sequential fashion. While AE systems are most needed to study properties that cannot be predicted analytically or computationally, even imperfect predictions can in principle be useful. Here, we use a case study on the mechanics of additively manufactured polymer structures to investigate whether imperfect data from simulation can accelerate AE. Initially, we study resilience, a property that is well-predicted by finite element analysis (FEA), and find that FEA can be used to build a Bayesian prior, and that experimental data can be integrated using discrepancy modeling to reduce the number of needed experiments ten-fold. Next, we study toughness, which is not well predicted by FEA, and find that FEA can still improve learning by transforming experimental data and guiding experiment selection. These results highlight multiple ways in which simulation can improve AE through transfer learning.

Current approaches to design efficient antisense RNAs (asRNAs) rely primarily on a thermodynamic understanding of RNA–RNA interactions. However, these approaches depend on structure predictions and have limited accuracy, arguably due to overlooking important cellular environment factors. In this work, we develop a biophysical model to describe asRNA–RNA hybridization that incorporates in vivo factors using large-scale experimental hybridization data for three model RNAs: a group I intron, CsrB and a tRNA. A unique element of our model is the estimation of the availability of the target region to interact with a given asRNA using a differential entropic consideration of suboptimal structures. We showcase the utility of this model by evaluating its prediction capabilities in four additional RNAs: a group II intron, Spinach II, 2-MS2 binding domain and glgC 5΄ UTR. Additionally, we demonstrate the applicability of this approach to other bacterial species by predicting sRNA–mRNA binding regions in two newly discovered, though uncharacterized, regulatory RNAs.

We present a technique for adaptively choosing a sequence of experiments for materials design and optimization. Specifically, we consider the problem of identifying the choice of experimental control variables that optimize the kinetic stability of a nanoemulsion, which we formulate as a ranking and selection problem. We introduce an optimization algorithm called the knowledge gradient with discrete priors (KGDP) that sequentially and adaptively selects experiments and that maximizes the rate of learning the optimal control variables. This is done through a combination of a physical, kinetic model of nanoemulsion stability, Bayesian inference, and a decision policy. Prior knowledge from domain experts is incorporated into the algorithm as well. Through numerical experiments, we show that the KGDP algorithm outperforms the policies of both random exploration (in which an experiment is selected uniformly at random among all potential experiments) and exploitation (which selects the experiment that appears t...

We consider the choices and subsequent costs associated with ensemble averaging and extrapolating experimental measurements in the context of optimizing material properties using Optimal Learning (OL). We demonstrate how these two general techniques lead to a trade-off between measurement error and experimental costs, and incorporate this trade-off in the OL framework. As a first contextual example, we study the effect of ensemble size in determining the most accessible regions of an RNA molecule. A second example considers the impact of the number and frequency of initial measurements used to extrapolate a measure of nanoemulsion stability. In both cases, we use OL simulations to determine the optimal choice of these characterization parameters by minimizing an associated total experimental cost.

We present a unified model of compound semiconductor growth based on kinetic Monte Carlo simulations in tandem with experimental results that can describe and predict the mechanisms for the formation of various types of nanostructures observed during droplet epitaxy. The crucial features of the model include the explicit and independent representation of atoms with different species and the ability to treat solid and liquid phases independently. Using this model, we examine nanostructural evolution in droplet epitaxy. The model faithfully captures several of the experimentally observed structures, including compact islands and nanorings. Moreover, simulations show the presence of Ga/GaAs core-shell structures that we validate experimentally. A fully analytical model of droplet epitaxy that explains the relationship between growth conditions and the resulting nanostructures is presented, yielding key insight into the mechanisms of droplet epitaxy. © 2013 American Physical Society.