Human versus Robots in the Discovery and Crystallization of Gigantic Polyoxometalates

Superlative mechanical energy absorbing efficiency discovered through self-driving lab-human partnership

Energy absorbing efficiency is a key determinant of a structure's ability to provide mechanical protection and is defined by the amount of energy that can be absorbed prior to stresses increasing to a level that damages the system to be protected. Here, we explore the energy absorbing efficiency of additively manufactured polymer structures by using a self-driving lab (SDL) to perform >25,000 physical experiments on generalized cylindrical shells. We use a human-SDL collaborative approach where experiments are selected from over trillions of candidates in an 11-dimensional parameter space using Bayesian optimization and then automatically performed while the human team monitors progress to periodically modify aspects of the system. The result of this human-SDL campaign is the discovery of a structure with a 75.2% energy absorbing efficiency and a library of experimental data that reveals transferable principles for designing tough structures.

In Pursuit of the Exceptional: Research Directions for Machine Learning in Chemical and Materials Science

Exceptional molecules and materials with one or more extraordinary properties are both technologically valuable and fundamentally interesting, because they often involve new physical phenomena or new compositions that defy expectations. Historically, exceptionality has been achieved through serendipity, but recently, machine learning (ML) and automated experimentation have been widely proposed to accelerate target identification and synthesis planning. In this Perspective, we argue that the data-driven methods commonly used today are well-suited for optimization but not for the realization of new exceptional materials or molecules. Finding such outliers should be possible using ML, but only by shifting away from using traditional ML approaches that tweak the composition, crystal structure, or reaction pathway. We highlight case studies of high-Tc oxide superconductors and superhard materials to demonstrate the challenges of ML-guided discovery and discuss the limitations of automation for this task. We then provide six recommendations for the development of ML methods capable of exceptional materials discovery: (i) Avoid the tyranny of the middle and focus on extrema; (ii) When data are limited, qualitative predictions that provide direction are more valuable than interpolative accuracy; (iii) Sample what can be made and how to make it and defer optimization; (iv) Create room (and look) for the unexpected while pursuing your goal; (v) Try to fill-in-the-blanks of input and output space; (vi) Do not confuse human understanding with model interpretability. We conclude with a description of how these recommendations can be integrated into automated discovery workflows, which should enable the discovery of exceptional molecules and materials.

Using Data-Driven Learning to Predict and Control the Outcomes of Inorganic Materials Synthesis

The design of inorganic materials for various applications critically depends on our ability to manipulate their synthesis in a rational, robust, and controllable fashion. Different from the conventional trial-and-error approach, data-driven techniques such as the design of experiments (DoE) and machine learning are an effective and more efficient way to predictably control materials synthesis. Here, we present a Viewpoint on recent progress in leveraging such techniques for predicting and controlling the outcomes of inorganic materials synthesis. We first compare how the design choice (statistical DoE vs machine learning) affects the type of control it can offer over the resulting product attributes, information elucidated, and experimental cost. These attributes are supported by discussing select case studies from the recent literature that highlight the power of these techniques for materials synthesis. The influence of experimental bias is next discussed, followed finally by our perspectives on the major challenges in the widespread implementation of predictable and controllable materials synthesis using data-driven techniques.

Robotic Modules for the Programmable Chemputation of Molecules and Materials

Before leveraging big data methods like machine learning and artificial intelligence (AI) in chemistry, there is an imperative need for an affordable, universal digitization standard. This mirrors the foundational requisites of the digital revolution, which demanded standard architectures with precise specifications. Recently, we have developed automated platforms tailored for chemical AI-driven exploration, including the synthesis of molecules, materials, nanomaterials, and formulations. Our focus has been on designing and constructing affordable standard hardware and software modules that serve as a blueprint for chemistry digitization across varied fields. Our platforms can be categorized into four types based on their applications: (i) discovery systems for the exploration of chemical space and novel reactivity, (ii) systems for the synthesis and manufacture of fine chemicals, (iii) platforms for formulation discovery and exploration, and (iv) systems for materials discovery and synthesis. We also highlight the convergent evolution of these platforms through shared hardware, firmware, and software alongside the creation of a unique programming language for chemical and material systems. This programming approach is essential for reliable synthesis, designing experiments, discovery, optimization, and establishing new collaboration standards. Furthermore, it is crucial for verifying literature findings, enhancing experimental outcome reliability, and fostering collaboration and sharing of unsuccessful experiments across different research labs.

Diversifying Design of Nucleic Acid Aptamers Using Unsupervised Machine Learning

Inverse design of short single-stranded RNA and DNA sequences (aptamers) is the task of finding sequences that satisfy a set of desired criteria. Relevant criteria may be, for example, the presence of specific folding motifs, binding to molecular ligands, sensing properties, and so on. Most practical approaches to aptamer design identify a small set of promising candidate sequences using high-throughput experiments (e.g., SELEX) and then optimize performance by introducing only minor modifications to the empirically found candidates. Sequences that possess the desired properties but differ drastically in chemical composition will add diversity to the search space and facilitate the discovery of useful nucleic acid aptamers. Systematic diversification protocols are needed. Here we propose to use an unsupervised machine learning model known as the Potts model to discover new, useful sequences with controllable sequence diversity. We start by training a Potts model using the maximum entropy principle on a small set of empirically identified sequences unified by a common feature. To generate new candidate sequences with a controllable degree of diversity, we take advantage of the model's spectral feature: an "energy" bandgap separating sequences that are similar to the training set from those that are distinct. By controlling the Potts energy range that is sampled, we generate sequences that are distinct from the training set yet still likely to have the encoded features. To demonstrate performance, we apply our approach to design diverse pools of sequences with specified secondary structure motifs in 30-mer RNA and DNA aptamers.

Exploring Optimal Reaction Conditions Guided by Graph Neural Networks and Bayesian Optimization

The optimization of organic reaction conditions to obtain the target product in high yield is crucial to avoid expensive and time-consuming chemical experiments. Advancements in artificial intelligence have enabled various data-driven approaches to predict suitable chemical reaction conditions. However, for many novel syntheses, the process to determine good reaction conditions is inevitable. Bayesian optimization (BO), an iterative optimization algorithm, demonstrates exceptional performance to identify reagents compared to synthesis experts. However, BO requires several initial randomly selected experimental results (yields) to train a surrogate model (approximately 10 experimental trials). Parts of this process, such as the cold-start problem in recommender systems, are inefficient. Here, we present an efficient optimization algorithm to determine suitable conditions based on BO that is guided by a graph neural network (GNN) trained on a million organic synthesis experiment data. The proposed method determined 8.0 and 8.7% faster high-yield reaction conditions than state-of-the-art algorithms and 50 human experts, respectively. In 22 additional optimization tests, the proposed method needed 4.7 trials on average to find conditions higher than the yield of the conditions recommended by five synthesis experts. The proposed method is considered in a situation of having a reaction dataset for training GNN.

Assemblies of Increasingly Large Ln-Containing Polyoxoniobates and Intermolecular Aggregation-Disaggregation Interconversions

An oxalate-assisted lanthanide (Ln) incorporation strategy is first demonstrated for creating rare high-nuclearity Ln-containing polyoxoniobates (PONbs). With the strategy, a series of high-nuclearity Ln-containing PONbs of 50-nuclearity Dy₂Nb₄₈, 103-nuclearity Dy₇Nb₉₆, 200-nuclearity Dy₁₀Nb₁₉₀, and 206-nuclearity Dy₁₄Nb₁₉₂ have been made, showing an increasingly large structure evolution from Dy₂Nb₄₈ monomer to Dy₇Nb₉₆ dimer and to distinct Dy₁₀Nb₁₉₀ and Dy₁₄Nb₁₉₂ tetramers. Among them, Dy₁₄Nb₁₉₂ presents the largest heterometallic PONb and also the PONb with the greatest number of Ln ions reported thus far. Interestingly, both giant Dy₁₄Nb₁₉₂ and Dy₁₀Nb₁₉₀ molecules can further undergo single-crystal to single-crystal intermolecular aggregations, forming infinite {Dy₁₄Nb₁₉₂}_∞ and {Dy₁₀Nb₁₉₀}_∞ chains, respectively. The former structural transformation shows a reversible humidity-dependent aggregation-disaggregation process accompanied by a proton conductivity response, while the latter structural transformation is irreversible. These new species largely enrich the very limited members of Ln-containing PONb family and offer rare examples for studying structural transformations between giant molecular aggregates and infinitely extended structures at the atomic level.

Data-driven efficient synthetic exploration of anionic lanthanide-based metal-organic frameworks

The synthesis of lanthanide metal-organic frameworks with terephthalate (Ln-BDC-MOFs) was investigated using a data-driven approach. Visually mapping the previously reported synthetic conditions suggested the existence of unexplored search spaces for novel Ln-BDC-MOFs. By focusing on the unexplored chemical reaction space, we successfully synthesized a series of new anionic Ln-BDC-MOFs, KGF-15, which demonstrated potential as luminescent sensors for Cu²⁺ ions. This synthetic exploration approach can significantly reduce the experimental effort required to discover new materials.

Lanthanides Singing the Blues: Their Fascinating Role in the Assembly of Gigantic Molybdenum Blue Wheels

Molybdenum blues (MBs) are a distinct class of polyoxometalates, exhibiting versatile/impressive architectures and high structural flexibility. In acidified and reduced aqueous environments, isopolymolybdates generate precisely organizable building blocks, which enable unique nanoscopic molecular systems (MBs) to be constructed and further fine-tuned by hetero elements such as lanthanide (Ln) ions. This Review discusses wheel-shaped MB-based structure types with strong emphasis on the ∼30 Ln-containing MBs as of August 2021, which include both organically hybridized and nonhybridized structures synthesized to date. The spotlight is thereby put on the lanthanide ions and ligand types, which are crucial for the resulting Ln-patterns and alterations in the gigantic structures. Several critical steps and reaction conditions in their synthesis are highlighted, as well as appropriate methods to investigate them both in solid state and in solution. The final section addresses the homogeneous/heterogeneous catalytic, molecular recognition and separation properties of wheel-shaped Ln-MBs, emphasizing their inimitable behavior and encouraging their application in these areas.

Evaluation guidelines for machine learning tools in the chemical sciences

Machine learning (ML) promises to tackle the grand challenges in chemistry and speed up the generation, improvement and/or ordering of research hypotheses. Despite the overarching applicability of ML workflows, one usually finds diverse evaluation study designs. The current heterogeneity in evaluation techniques and metrics leads to difficulty in (or the impossibility of) comparing and assessing the relevance of new algorithms. Ultimately, this may delay the digitalization of chemistry at scale and confuse method developers, experimentalists, reviewers and journal editors. In this Perspective, we critically discuss a set of method development and evaluation guidelines for different types of ML-based publications, emphasizing supervised learning. We provide a diverse collection of examples from various authors and disciplines in chemistry. While taking into account varying accessibility across research groups, our recommendations focus on reporting completeness and standardizing comparisons between tools. We aim to further contribute to improved ML transparency and credibility by suggesting a checklist of retro-/prospective tests and dissecting their importance. We envisage that the wide adoption and continuous update of best practices will encourage an informed use of ML on real-world problems related to the chemical sciences.

Synthesis and characterization of the `Japanese rice-ball'-shaped Molybdenum Blue Na4[Mo2O2(OH)4(C6H4NO2)2]2[Mo120Ce6O366H12(OH)2(H2O)76]∼200H2O

The synthesis and crystal structure of a ‘Japanese rice-ball’-shaped Molybdenum Blue hybridized organically with 2-picolinic acid is presented. In addition to single-crystal X-ray analysis, the title com­pound was characterized with IR spectroscopy and elemental analyses to reinforce its framework structure, thermogravimetric analysis to qu­antify the water content, and Vis–NIR spectroscopy to determine the degree of reduction of the nanocluster.

The hybridized lanthanide-containing molybdenum blue (Ln-MB) wheel Na₄[Mo₂O₂(OH)₄(C₆H₄NO₂)₂]₂[Mo₁₂₀Ce₆O₃₆₆H₁₂(OH)₂(H₂O)₇₆]∼200H₂O ({Mo₂(C₆H₄NO₂)₂}₂{Mo₁₂₀Ce₆}) was assembled in an aqueous one-pot synthesis. The Ln-MB was hybridized with 2-picolinic acid through the generation of the organometallic counter-ion [Mo₂O₂(OH)₄(C₆H₄NO₂)₂]²⁺. Control experiments demonstrated that the position of the carb­oxy­lic acid group (2-position to the N atom) in the hybridization com­ponent is critical in yielding single crystals of Ln-MB. In addition to single-crystal X-ray diffraction (XRD) analysis, which revealed a ‘Japanese rice-ball’-shaped Ln-MB as the anion, elemental analyses, IR spectroscopy, and thermogravimetric analysis (TGA) were performed to confirm its structure and com­position. Bond-valence-sum calculations (BVS) revealed that {Mo₂(C₆H₄NO₂)₂}₂{Mo₁₂₀Ce₆} is com­posed of a 24-electron re­duced anionic ring, which was confirmed by Vis–NIR spectroscopy.

Labels in a haystack: Approaches beyond supervised learning in biomedical applications

Recent advances in biomedical machine learning demonstrate great potential for data-driven techniques in health care and biomedical research. However, this potential has thus far been hampered by both the scarcity of annotated data in the biomedical domain and the diversity of the domain's subfields. While unsupervised learning is capable of finding unknown patterns in the data by design, supervised learning requires human annotation to achieve the desired performance through training. With the latter performing vastly better than the former, the need for annotated datasets is high, but they are costly and laborious to obtain. This review explores a family of approaches existing between the supervised and the unsupervised problem setting. The goal of these algorithms is to make more efficient use of the available labeled data. The advantages and limitations of each approach are addressed and perspectives are provided.

Failure-Experiment-Supported Optimization of Poorly Reproducible Synthetic Conditions for Novel Lanthanide Metal-Organic Frameworks with Two-Dimensional Secondary Building Units*

Novel metal-organic frameworks containing lanthanide double-layer-based secondary building units (KGF-3) were synthesized by using machine learning (ML). Isolating pure KGF-3 was challenging, and the synthesis was not reproducible because impurity phases were frequently obtained under the same synthetic conditions. Thus, dominant factors for the synthesis of KGF-3 were identified, and its synthetic conditions were optimized by using two ML techniques. Cluster analysis was used to classify the obtained powder X-ray diffractometry patterns of the products and thus automatically determine whether the experiments were successful. Decision-tree analysis was used to visualize the experimental results, after extracting factors that mainly affected the synthetic reproducibility. Water-adsorption isotherms revealed that KGF-3 possesses unique hydrophilic pores. Impedance measurements demonstrated good proton conductivities (σ=5.2×10^-4  S cm^-1 for KGF-3(Y)) at a high temperature (363 K) and relative humidity of 95 % RH.

Assisted Self-Assembly to Target Heterometallic Mn-Nd and Mn-Sm SMMs: Synthesis and Magnetic Characterisation of [Mn7 Ln3 (O)4 (OH)4 (mdea)3 (piv)9 (NO3 )3 ] (Ln=Nd, Sm, Eu, Gd)*

In an assisted self-assembly approach starting from the [Mn6 O2 (piv)10 (4-Me-py)2 (pivH)2 ] cluster a family of Mn-Ln compounds (Ln=Pr-Yb) was synthesised. The reaction of [Mn6 O2 (piv)10 (4-Me-py)2 (pivH)2 ] (1) with N-methyldiethanolamine (mdeaH2 ) and Ln(NO3 )3  ⋅ 6H2 O in MeCN generally yields two main structure types: for Ln=Tb-Yb a previously reported Mn5 Ln4 motif is obtained, whereas for Ln=Pr-Eu a series of Mn7 Ln3 clusters is obtained. Within this series the GdIII analogue represents a special case because it shows both structural types as well as a third Mn2 Ln2 inverse butterfly motif. Variation in reaction conditions allows access to different structure types across the whole series. This prompts further studies into the reaction mechanism of this cluster assisted self-assembly approach. For the Mn7 Ln3 analogues reported here variable-temperature magnetic susceptibility measurements suggest that antiferromagnetic interactions between the spin carriers are dominant. Compounds incorporating Ln=NdIII (2), SmIII (3) and GdIII (5) display SMM behaviour. The slow relaxation of the magnetisation for these compounds was confirmed by ac measurements above 1.8 K.

Chemputation and the Standardization of Chemical Informatics

The explosion in the use of machine learning for automated chemical reaction optimization is gathering pace. However, the lack of a standard architecture that connects the concept of chemical transformations universally to software and hardware provides a barrier to using the results of these optimizations and could cause the loss of relevant data and prevent reactions from being reproducible or unexpected findings verifiable or explainable. In this Perspective, we describe how the development of the field of digital chemistry or chemputation, that is the universal code-enabled control of chemical reactions using a standard language and ontology, will remove these barriers allowing users to focus on the chemistry and plug in algorithms according to the problem space to be explored or unit function to be optimized. We describe a standard hardware (the chemical processing programming architecture-the ChemPU) to encompass all chemical synthesis, an approach which unifies all chemistry automation strategies, from solid-phase peptide synthesis, to HTE flow chemistry platforms, while at the same time establishing a publication standard so that researchers can exchange chemical code (χDL) to ensure reproducibility and interoperability. Not only can a vast range of different chemistries be plugged into the hardware, but the ever-expanding developments in software and algorithms can also be accommodated. These technologies, when combined will allow chemistry, or chemputation, to follow computation-that is the running of code across many different types of capable hardware to get the same result every time with a low error rate.

Robotic Stepwise Synthesis of Hetero-Multinuclear Metal Oxo Clusters as Single-Molecule Magnets

An efficient stepwise synthesis method for discovering new heteromultinuclear metal clusters using a robotic workflow is developed where numerous reaction conditions for constructing heteromultinuclear metal oxo clusters in polyoxometalates (POMs) were explored using a custom-built automated platform. As a result, new nonanuclear tetrametallic oxo clusters {FeMn₄}Lu₂A₂ in TBA₅[(A-α-SiW₉O₃₄)₂FeMn₄O₂{Lu(acac)₂}₂A₂] (II^A; A = Ag, Na, K; TBA = tetra-n-butylammonium; acac = acetylacetonate) were discovered by the installation of diamagnetic metal cations A⁺ into a paramagnetic {FeMn₄}Lu₂ unit in TBA₇[(A-α-SiW₉O₃₄)₂FeMn₄O₂{Lu(acac)₂}₂] (I). POMs II^A exhibited single-molecule magnet properties with the higher energy barriers for magnetization reversal (II^Ag, 40.0 K; II^Na, 40.3 K; II^K, 26.7 K) compared with that of the parent I (19.7 K). Importantly, these clusters with unique properties were constructed as designed by a step of the predictable sequential multistep reactions with the time-efficient platform.

Self-Driving Laboratories for Development of New Functional Materials and Optimizing Known Reactions

Innovations often play an essential role in the acceleration of the new functional materials discovery. The success and applicability of the synthesis results with new chemical compounds and materials largely depend on the previous experience of the researcher himself and the modernity of the equipment used in the laboratory. Artificial intelligence (AI) technologies are the next step in developing the solution for practical problems in science, including the development of new materials. Those technologies go broadly beyond the borders of a computer science branch and give new insights and practical possibilities within the far areas of expertise and chemistry applications. One of the attractive challenges is an automated new functional material synthesis driven by AI. However, while having many years of hands-on experience, chemistry specialists have a vague picture of AI. To strengthen and underline AI's role in materials discovery, a short introduction is given to the essential technologies, and the machine learning process is explained. After this review, this review summarizes the recent studies of new strategies that help automate and accelerate the development of new functional materials. Moreover, automatized laboratories' self-driving cycle could benefit from using AI algorithms to optimize new functional nanomaterials' synthetic routes. Despite the fact that such technologies will shape material science in the nearest future, we note the intelligent use of algorithms and automation is required for novel discoveries.

ChemOS: Orchestrating autonomous experimentation

ChemOS aims to catalyze the expansion of autonomous laboratories and to disrupt the conventional approach to experimentation.

Integration of metallacycles and polyoxometalate macrocycles

A series of polyoxometalate–metallacycle composite macrocycles with interesting bi/tri-layer cyclic structures have been synthesized, which show a remarkable integration of metallacycles and polyoxometalate macrocycles at molecular level.

Regulating the solution structural integrity and slow magnetic relaxation behavior of two Dy6 clusters with a pyridine–triazole ligand

Two {Dy₆} skeletons consolidated by nitrate (1) and hydroxide (2) have been isolated. HRESI-MS reveals that 1 retains a higher structural integrity and fragment stability. While 2 has a typical slow magnetic relaxation of SMM behavior.

Peptide sequence mediated self-assembly of molybdenum blue nanowheel superstructures

The precise control over the formation of complex nanostructures, e.g. polyoxometalates (POMs), at the sub-nanoscale is challenging but critical if non-covalent architectures are to be designed. Combining biologically-evolved systems with inorganic nanostructures could lead to sequence-mediated assembly. Herein, we exploit oligopeptides as multidentate structure-directing ligands via metal-coordination and hydrogen bonded interactions to modulate the self-assembly of POM superstructures. Six oligopeptides (GH, AH, SH, G₂H, G₄H and G₅H) are incorporated into the cavities of Molybdenum Blue (MB) POM nanowheels. It is found that the helicity of the nanowheel can be readily switched (Δ to Λ) by simply altering the N-terminal amino acid on the peptide chain rather than their overall stereochemistry. We also reveal a delicate balance between the Mo-coordination and the hydrogen bonds found within the internal cavity of the inorganic nanowheels which results in the sequence mediated formation of two unprecedented asymmetrical nanowheel frameworks: {Mo₁₂₂Ce₅} and {Mo₁₂₆Ce₄}.

Peptide sequence can be used to control the self-assembly and structures of nanoscale molybdenum blue polyoxometalate (POM) wheel-shaped clusters.

Statistical Analysis of Glycosylation Reactions

Chemical synthesis is one of the practical approaches to access carbohydrate-based natural products and their derivatives with high quality and in a large quantity. However, stereoselectivity during the glycosylation reaction is the main challenge because the reaction can generate both α- and β-glycosides. The main focus of the present article is the concept of recent mechanistic studies that have applied statistical analysis and quantitation for defining stereoselective changes during the reaction process. Based on experimental evidence, a detailed discussion associated with the mechanism and degree of influence affecting the stereoselective outcome of glycosylation is included.

A Modular Programmable Inorganic Cluster Discovery Robot for the Discovery and Synthesis of Polyoxometalates

The exploration of complex multicomponent chemical reactions leading to new clusters, where discovery requires both molecular self-assembly and crystallization, is a major challenge. This is because the systematic approach required for an experimental search is limited when the number of parameters in a chemical space becomes too large, restricting both exploration and reproducibility. Herein, we present a synthetic strategy to systematically search a very large set of potential reactions, using an inexpensive, high-throughput platform that is modular in terms of both hardware and software and is capable of running multiple reactions with in-line analysis, for the automation of inorganic and materials chemistry. The platform has been used to explore several inorganic chemical spaces to discover new and reproduce known tungsten-based, mixed transition-metal polyoxometalate clusters, giving a digital code that allows the easy repeat synthesis of the clusters. Among the many species identified in this work, the most significant is the discovery of a novel, purely inorganic W₂₄Fe^III-superoxide cluster formed under ambient conditions. The modular wheel platform was employed to undertake two chemical space explorations, producing compounds 1-4: (C₂H₈N)₁₀Na₂[H₆Fe(O₂)W₂₄O₈₂] (1, {W₂₄Fe}), (C₂H₈N)₇₂Na₁₆[H₁₆Co₈W₂₀₀O₆₆₀(H₂O)₄₀] (2, {W₂₀₀Co₈}), (C₂H₈N)₇₂Na₁₆[H₁₆Ni₈W₂₀₀O₆₆₀(H₂O)₄₀] (3, {W₂₀₀Ni₈}), and (C₂H₈N)₁₄[H₂₆W₃₄V₄O₁₃₀] (4, {W₃₄V₄}), along with many other known species, such as simple Keggin clusters and 1D {W₁₁M²⁺} chains.

Molecular Machine Learning: The Future of Synthetic Chemistry?

During the last decade, modern machine learning has found its way into synthetic chemistry. Some long-standing challenges, such as computer-aided synthesis planning (CASP), have been successfully addressed, while other issues have barely been touched. This Viewpoint poses the question of whether current trends can persist in the long term and identifies factors that may lead to an (un)productive development. Thereby, specific risks of molecular machine learning (MML) are discussed. Furthermore, possible sustainable developments are suggested, such as explainable artificial intelligence (exAI) for synthetic chemistry. This Viewpoint will illuminate chances for possible newcomers and aims to guide the community into a discussion about current as well as future trends.

High-Throughput Approaches for the Discovery of Supramolecular Organic Cages

The assembly of complex molecules, such as organic cages, can be achieved through supramolecular and dynamic covalent strategies. Their use in a range of applications has been demonstrated, including gas uptake, molecular separations, and in catalysis. However, the targeted design and synthesis of new species for particular applications is challenging, particularly as the systems become more complex. High-throughput computation-only and experiment-only approaches have been developed to streamline the discovery process, although are still not widely implemented. Additionally, combined hybrid workflows can dramatically accelerate the discovery process and lead to the serendipitous discovery and rationalisation of new supramolecular assemblies that would not have been designed based on intuition alone. This Minireview focuses on the advances in high-throughput approaches that have been developed and applied in the discovery of supramolecular organic cages.

Nucleation mechanisms and speciation of metal oxide clusters

The self-assembly mechanisms of polyoxometalates (POMs) are still a matter of discussion owing to the difficult task of identifying all the chemical species and reactions involved. We present a new computational methodology that identifies the reaction mechanism for the formation of metal-oxide clusters and provides a speciation model from first-principles and in an automated manner. As a first example, we apply our method to the formation of octamolybdate. In our model, we include variables such as pH, temperature and ionic force because they have a determining effect on driving the reaction to a specific product. Making use of graphs, we set up and solved 2.8 × 105 multi-species chemical equilibrium (MSCE) non-linear equations and found which set of reactions fitted best with the experimental data available. The agreement between computed and experimental speciation diagrams is excellent. Furthermore, we discovered a strong linear dependence between DFT and empirical formation constants, which opens the door for a systematic rescaling.

Practical considerations for active machine learning in drug discovery

Active machine learning enables the automated selection of the most valuable next experiments to improve predictive modelling and hasten active retrieval in drug discovery. Although a long established theoretical concept and introduced to drug discovery approximately 15 years ago, the deployment of active learning technology in the discovery pipelines across academia and industry remains slow. With the recent re-discovered enthusiasm for artificial intelligence as well as improved flexibility of laboratory automation, active learning is expected to surge and become a key technology for molecular optimizations. This review recapitulates key findings from previous active learning studies to highlight the challenges and opportunities of applying adaptive machine learning to drug discovery. Specifically, considerations regarding implementation, infrastructural integration, and expected benefits are discussed. By focusing on these practical aspects of active learning, this review aims at providing insights for scientists planning to implement active learning workflows in their discovery pipelines.

An Autonomous Chemical Robot Discovers the Rules of Inorganic Coordination Chemistry without Prior Knowledge

We present a chemical discovery robot for the efficient and reliable discovery of supramolecular architectures through the exploration of a huge reaction space exceeding ten billion combinations. The system was designed to search for areas of reactivity found through autonomous selection of the reagent types, amounts, and reaction conditions aiming for combinations that are reactive. The process consists of two parts where reagents are mixed together, choosing from one type of aldehyde, one amine and one azide (from a possible family of two amines, two aldehydes and four azides) with different volumes, ratios, reaction times, and temperatures, whereby the reagents are passed through a copper coil reactor. Next, either cobalt or iron is added, again from a large number of possible quantities. The reactivity was determined by evaluating differences in pH, UV‐Vis, and mass spectra before and after the search was started. The algorithm was focused on the exploration of interesting regions, as defined by the outputs from the sensors, and this led to the discovery of a range of 1‐benzyl‐(1,2,3‐triazol‐4‐yl)‐N‐alkyl‐(2‐pyridinemethanimine) ligands and new complexes: [Fe(L¹)₂](ClO₄)₂ (1); [Fe(L²)₂](ClO₄)₂ (2); [Co₂(L³)₂](ClO₄)₄ (3); [Fe₂(L³)₂](ClO₄)₄ (4), which were crystallised and their structure confirmed by single‐crystal X‐ray diffraction determination, as well as a range of new supramolecular clusters discovered in solution using high‐resolution mass spectrometry.

ChemOS: An orchestration software to democratize autonomous discovery

The current Edisonian approach to discovery requires up to two decades of fundamental and applied research for materials technologies to reach the market. Such a slow and capital-intensive turnaround calls for disruptive strategies to expedite innovation. Self-driving laboratories have the potential to provide the means to revolutionize experimentation by empowering automation with artificial intelligence to enable autonomous discovery. However, the lack of adequate software solutions significantly impedes the development of self-driving laboratories. In this paper, we make progress towards addressing this challenge, and we propose and develop an implementation of ChemOS; a portable, modular and versatile software package which supplies the structured layers necessary for the deployment and operation of self-driving laboratories. ChemOS facilitates the integration of automated equipment, and it enables remote control of automated laboratories. ChemOS can operate at various degrees of autonomy; from fully unsupervised experimentation to actively including inputs and feedbacks from researchers into the experimentation loop. The flexibility of ChemOS provides a broad range of functionality as demonstrated on five applications, which were executed on different automated equipment, highlighting various aspects of the software package.

A novel sandwich shaped {CoIII2CoII12MoV24} cluster with a CoII4 triangle encapsulated in two capped CoIIICoII4MoV12O40 fragments

A novel discrete {Co₁₄Mo₂₄} nanoscaled cluster, with a triangle Co₄ core encapsulated in two novel capped Co-substituted Keggin-type Co₅Mo₁₂O₄₀ anions, was isolated from alkaline methanol solution. And the HRESI-MS of {Co₁₄Mo₂₄} cluster was recorded.

A curious formulation robot enables the discovery of a novel protocell behavior

We describe a chemical robotic assistant equipped with a curiosity algorithm (CA) that can efficiently explore the states a complex chemical system can exhibit. The CA-robot is designed to explore formulations in an open-ended way with no explicit optimization target. By applying the CA-robot to the study of self-propelling multicomponent oil-in-water protocell droplets, we are able to observe an order of magnitude more variety in droplet behaviors than possible with a random parameter search and given the same budget. We demonstrate that the CA-robot enabled the observation of a sudden and highly specific response of droplets to slight temperature changes. Six modes of self-propelled droplet motion were identified and classified using a time-temperature phase diagram and probed using a variety of techniques including NMR. This work illustrates how CAs can make better use of a limited experimental budget and significantly increase the rate of unpredictable observations, leading to new discoveries with potential applications in formulation chemistry.

The stepwise substitution in the hierarchical building of {Co11Cd6} cluster-based MOFs from {Co14} precursor

A heterometallic aggregation of {Co₁₁Cd₆} can be hierarchical built from a {Co₁₄} precursor directly. HRESI-MS technique was used to track the stepwise Cd²⁺-capture: {Co₁₄} → {Co₈} → {Co₈Cd₂} → {Co₈Cd₃} → {Co₈Cd₄} → {Co₈Cd₅} → {Co₈Cd₆} → {Co₁₁Cd₆}.

Machine learning for target discovery in drug development

The discovery of macromolecular targets for bioactive agents is currently a bottleneck for the informed design of chemical probes and drug leads. Typically, activity profiling against genetically manipulated cell lines or chemical proteomics is pursued to shed light on their biology and deconvolute drug-target networks. By taking advantage of the ever-growing wealth of publicly available bioactivity data, learning algorithms now provide an attractive means to generate statistically motivated research hypotheses and thereby prioritize biochemical screens. Here, we highlight recent successes in machine intelligence for target identification and discuss challenges and opportunities for drug discovery.

Anthropogenic biases in chemical reaction data hinder exploratory inorganic synthesis

Most chemical experiments are planned by human scientists and therefore are subject to a variety of human cognitive biases¹, heuristics² and social influences³. These anthropogenic chemical reaction data are widely used to train machine-learning models⁴ that are used to predict organic⁵ and inorganic^6,7 syntheses. However, it is known that societal biases are encoded in datasets and are perpetuated in machine-learning models⁸. Here we identify as-yet-unacknowledged anthropogenic biases in both the reagent choices and reaction conditions of chemical reaction datasets using a combination of data mining and experiments. We find that the amine choices in the reported crystal structures of hydrothermal synthesis of amine-templated metal oxides⁹ follow a power-law distribution in which 17% of amine reactants occur in 79% of reported compounds, consistent with distributions in social influence models^10-12. An analysis of unpublished historical laboratory notebook records shows similarly biased distributions of reaction condition choices. By performing 548 randomly generated experiments, we demonstrate that the popularity of reactants or the choices of reaction conditions are uncorrelated to the success of the reaction. We show that randomly generated experiments better illustrate the range of parameter choices that are compatible with crystal formation. Machine-learning models that we train on a smaller randomized reaction dataset outperform models trained on larger human-selected reaction datasets, demonstrating the importance of identifying and addressing anthropogenic biases in scientific data.

A [13]rotaxane assembled via a palladium molecular capsule

Molecules that are the size of small proteins are difficult to make. The most frequently examined route is via self-assembly, and one particular approach involves molecular nanocapsules, where ligands are designed that will enforce the formation of specific polyhedra of metals within the core of the structure. Here we show that this approach can be combined with mechanically interlocking molecules to produce nanocapsules that are decorated on their exterior. This could be a general route to very large molecules, and is exemplified here by the synthesis and structural characterization of a [13]rotaxane, containing 150 metal centres. Small angle X-ray scattering combined with atomistic molecular dynamics simulations demonstrate the compound is intact in solution.

Mechanically interlocked molecules and molecular cages are two important themes in supramolecular chemistry. Here, the authors combine these concepts to construct a giant [13]rotaxane built around a palladium capsule, one of the most complex metallosupramolecular assemblies yet.

Machine learning and molecular descriptors enable rational solvent selection in asymmetric catalysis

Rational solvent selection remains a significant challenge in process development. Here we describe a hybrid mechanistic-machine learning approach, geared towards automated process development workflow. A library of 459 solvents was used, for which 12 conventional molecular descriptors, two reaction-specific descriptors, and additional descriptors based on screening charge density, were calculated. Gaussian process surrogate models were trained on experimental data from a Rh(CO)2(acac)/Josiphos catalysed asymmetric hydrogenation of a chiral α-β unsaturated γ-lactam. With two simultaneous objectives - high conversion and high diastereomeric excess - the multi-objective algorithm, trained on the initial dataset of 25 solvents, has identified solvents leading to better reaction outcomes. In addition to being a powerful design of experiments (DoE) methodology, the resulting Gaussian process surrogate model for conversion is, in statistical terms, predictive, with a cross-validation correlation coefficient of 0.84. After identifying promising solvents, the composition of solvent mixtures and optimal reaction temperature were found using a black-box Bayesian optimisation. We then demonstrated the application of a new genetic programming approach to select an appropriate machine learning model for a specific physical system, which should allow the transition of the overall process development workflow into the future robotic laboratories.

Intuition-Enabled Machine Learning Beats the Competition When Joint Human-Robot Teams Perform Inorganic Chemical Experiments

Traditionally, chemists have relied on years of training and accumulated experience in order to discover new molecules. But the space of possible molecules is so vast that only a limited exploration with the traditional methods can be ever possible. This means that many opportunities for the discovery of interesting phenomena have been missed, and in addition, the inherent variability of these phenomena can make them difficult to control and understand. The current state-of-the-art is moving toward the development of automated and eventually fully autonomous systems coupled with in-line analytics and decision-making algorithms. Yet even these, despite the substantial progress achieved recently, still cannot easily tackle large combinatorial spaces, as they are limited by the lack of high-quality data. Herein, we explore the utility of active learning methods for exploring the chemical space by comparing the collaboration between human experimenters with an algorithm-based search against their performance individually to probe the self-assembly and crystallization of the polyoxometalate cluster Na₆[Mo₁₂₀Ce₆O₃₆₆H₁₂(H₂O)₇₈]·200H₂O (1). We show that the robot-human teams are able to increase the prediction accuracy to 75.6 ± 1.8%, from 71.8 ± 0.3% with the algorithm alone and 66.3 ± 1.8% from only the human experimenters demonstrating that human-robot teams can beat robots or humans working alone.