Analogous to the way the Human Genome Project advanced an array of biological sciences by mapping the human genome, the Materials Genome Initiative aims to enhance our understanding of the fundamentals of materials science by providing the information we need to accelerate the development of new materials.This approach is particularly applicable to recently developed classes of nanoporous materials, such as metal-organic frameworks (MOFs), which are synthesized from a limited set of molecular building blocks that can be combined to generate a very large number of different structures. In this Perspective, we illustrate how a materials genome approach can be used to search for high-performance adsorbent materials to store natural gas in a vehicular fuel tank. Drawing upon recent reports of large databases of existing and predicted nanoporous materials generated in silico, we have collected and compared on a consistent basis the methane uptake in over 650 000 materials based on the results of molecular simulation. The data that we have collected provide candidate structures for synthesis, reveal relationships between structural characteristics and performance, and suggest that it may be difficult to reach the current Advanced Research Project Agency-Energy (ARPA-E) target for natural gas storage. Broader contextNatural gas, mostly methane, is an attractive replacement for petroleum fuels for automotive vehicles because of its economic and environmental advantages. However, it suffers from a low volumetric energy density, necessitating densication to yield reasonable driving ranges from a full fuel tank. Densication strategies in the market today, liquefaction and compression, require expensive and cumbersome vehicular fuel tanks and rell station infrastructure. If we are able to develop a nanoporous adsorbent material to store natural gas at ambient temperature and moderate pressure, one could envision a simple fuel tank that can be relled at home. Modern, advanced nanoporous materials are highly tunable, inundating researchers with practically innite possibilities of materials to synthesize and test for methane storage. The current research is focused on nding among these millions of possible materials one that can be used to store natural gas without using liquefaction or compression processes. In this Perspective, we adopt a computational approach to screen databases of over 650 000 nanoporous material structures. Using this nanoporous materials genome approach, we reveal relationships between structural characteristics and performance, and suggest that it may be difficult, if not impossible, to reach the current Advanced Research Project Agency-Energy (ARPA-E) target for natural gas storage using nanoporous materials.

Accelerating progress in the discovery and deployment of advanced nanoporous materials relies on chemical insight and structure property relationships for rational design. Because of the complexity of this problem, trial-and-error is heavily involved in the laboratory today. A cost-effective route to aid experimental materials discovery is to construct structure models of nanoporous materials in silico and use molecular simulations to rapidly test them and elucidate data-driven guidelines for rational design. For example, highly-tunable nanoporous materials have shown promise as adsorbents for separating an industrially relevant gaseous mixture of xenon and krypton. In this work, we characterize, screen, and analyze the Nanoporous Materials Genome, a database of ca. 670,000 porous material structures, for candidate adsorbents for xenon/krypton separations. For over half a million structures, the computational resources required for a brute-force screening using grand-canonical Monte Carlo simulations of Xe/Kr adsorption are prohibitive.To overcome the computational cost, we used a hybrid approach combining machine learning algorithms (random forests) with molecular simulations. We compared the results from our large-scale screening with simple pore models to rationalize the strong link between pore size and selectivity. With this insight, we then analyzed the anatomy of the binding sites of the most selective materials. These binding sites can be constructed from tubes, pockets, rings, or cages and are often composed of non-discrete chemical fragments. The complexity of these binding sites emphasizes the importance of high-throughput computational screenings to discover new materials. Interestingly, our screening study predicts that the two most selective materials in the database are an aluminophosphate zeolite analogue and a calcium based coordination network, both of which have already been synthesized but not yet tested for Xe/Kr separations.

The technological advances of the past century, marked by the computer revolution and the advent of high-throughput screening technologies in drug discovery, opened the path to the computational analysis and visualization of bioactive molecules. For this purpose, it became necessary to represent molecules in a syntax that would be readable by computers and understandable by scientists of various fields. A large number of chemical representations have been developed over the years, their numerosity being due to the fast development of computers and the complexity of producing a representation that encompasses all structural and chemical characteristics. We present here some of the most popular electronic molecular and macromolecular representations used in drug discovery, many of which are based on graph representations. Furthermore, we describe applications of these representations in AI-driven drug discovery. Our aim is to provide a brief guide on structural representations that are essential to the practice of AI in drug discovery. This review serves as a guide for researchers who have little experience with the handling of chemical representations and plan to work on applications at the interface of these fields.

We present accurate force fields developed from density functional theory (DFT) calculations with periodic boundary conditions for use in molecular simulations involving M2(dobdc) (M-MOF-74; dobdc4– = 2,5-dioxidobenzenedicarboxylate; M = Mg, Mn, Fe, Co, Ni, Zn) and frameworks of similar topology. In these systems, conventional force fields fail to accurately model gas adsorption due to the strongly binding open-metal sites. The DFT-derived force fields predict the adsorption of CO2, H2O, and CH4 inside these frameworks much more accurately than other common force fields. We show that these force fields can also be used for M2(dobpdc) (dobpdc4– = 4,4′-dioxidobiphenyl-3,3′-dicarboxylate), an extended version of MOF-74, and thus are a promising alternative to common force fields for studying materials similar to MOF-74 for carbon capture applications. Furthermore, it is anticipated that the approach can be applied to other metal–organic framework topologies to obtain force fields for different systems. We have used this force field to study the effect of contaminants such as H2O and N2 upon these materials’ performance for the separation of CO2 from the emissions of natural gas reservoirs and coal-fired power plants. Specifically, mixture adsorption isotherms calculated with these DFT-derived force fields showed a significant reduction in the uptake of many gas components in the presence of even trace amounts of H2O vapor. The extent to which the various gases are affected by the concentration of H2O in the reservoir is quantitatively different for the different frameworks and is related to their heats of adsorption. Additionally, significant increases in CO2 selectivities over CH4 and N2 are observed as the temperature of the systems is lowered.

Here, we present a database of 69 840 largely novel covalent organic frameworks assembled in silico from 666 distinct organic linkers and four established synthetic routes. Due to their light weights and high internal surface areas, the frameworks are promising materials for methane storage applications. To assess their methane storage performance, we used grand-canonical Monte Carlo simulations to calculate their deliverable capacities. We demonstrate that the best structure, composed of carbon− carbon bonded triazine linkers in the tbd topology, has a predicted 65-bar deliverable capacity of 216 v STP/v, better than the best methane storage materials published to date. Using our approach, we also discovered other high-performing materials with 300 structures having calculated deliverable capacities greater than 190 v STP/v and 10% of these outperforming 200 v STP/v. To encourage screening studies of these materials for other applications, all structures and their properties were made available on the Materials Cloud.

SignificanceNanocarbons can be characterized by their curvature—that is, positively curved fullerenes, zero-curved graphene, and negatively curved schwarzites. Schwartzites are fascinating materials but have not been synthesized yet, although disordered materials with local properties similar to schwarzites (“random schwarzites”) have been isolated. A promising synthetic method allows for the interior surfaces of zeolites to be templated with normalsnormalp2 carbon, but theoretical study of these zeolite-templated carbons (ZTCs) has been limited because of their noncrystalline structures. In this work, we develop an improved molecular description of ZTCs, show that they are equivalent to schwartzites, and thus make the experimental discovery of schwarzites ex post facto. Our topological characterization of ZTCs lends insights into how template choice allows for the tunability of ordered microporous carbons.

Porous covalent polymers are attracting increasing interest in the fields of gas adsorption, gas separation, and catalysis due to their fertile synthetic polymer chemistry, large internal surface areas, and ultrahigh hydrothermal stabilities. While precisely manipulating the porosities of porous organic materials for targeted applications remains challenging, we show how a large degree of diversity can be achieved in covalent organic polymers by incorporating multiple functionalities into a single framework, as is done for crystalline porous materials. Here, we synthesized 17 novel porous covalent organic polymers (COPs) with finely tuned porosities, a wide range of Brunauer-Emmett-Teller (BET) specific surface areas of 430-3624 m(2) g(-1), and a broad range of pore volumes of 0.24-3.50 cm(3) g(-1), all achieved by tailoring the length and geometry of building blocks. Furthermore, we are the first to successfully incorporate more than three distinct functional groups into one phase for porous organic materials, which has been previously demonstrated in crystalline metal-organic frameworks (MOFs). COPs decorated with multiple functional groups in one phase can lead to enhanced properties that are not simply linear combinations of the pure component properties. For instance, in the dibromobenzene-lined frameworks, the bi- and multifunctionalized COPs exhibit selectivities for carbon dioxide over nitrogen twice as large as any of the singly functionalized COPs. These multifunctionalized frameworks also exhibit a lower parasitic energy cost for carbon capture at typical flue gas conditions than any of the singly functionalized frameworks. Despite the significant improvement, these frameworks do not yet outperform the current state-of-art technology for carbon capture. Nonetheless, the tuning strategy presented here opens up avenues for the design of novel catalysts, the synthesis of functional sensors from these materials, and the improvement in the performance of existing covalent organic polymers by multifunctionalization.

Deep learning methods applied to chemistry can be used to accelerate the discovery of new molecules. This work introduces GraphINVENT, a platform developed for graph-based molecular design using graph neural networks (GNNs). GraphINVENT uses a tiered deep neural network architecture to probabilistically generate new molecules a single bond at a time. All models implemented in GraphINVENT can quickly learn to build molecules resembling the training set molecules without any explicit programming of chemical rules. The models have been benchmarked using the MOSES distribution-based metrics, showing how GraphINVENT models compare well with state-of-the-art generative models. This work compares six different GNN-based generative models in GraphINVENT, and shows that ultimately the gated-graph neural network performs best against the metrics considered here.