Computational compound screening of biomolecules and soft materials by molecular simulations

Abstract

Decades of hardware, methodological, and algorithmic development have propelled molecular dynamics (MD) simulations to the forefront of materialsmodeling techniques, bridging the gap between electronic-structure theory and continuum methods. The physics-based approach makes MD appropriate to study emergent phenomena, but simultaneously incurs significant computational investment. This topical review explores the use of MD outside the scope of individual systems, but rather considering many compounds. Such an in silico screening approach makes MD amenable to establishing coveted structure-property relationships. We specifically focus on biomolecules and soft materials, characterized by the significant role of entropic contributions and heterogeneous systems and scales. An account of the state of the art for the implementation of an MD-based screening paradigm is described, including automated force-field parametrization, system preparation, and efficient sampling across both conformation and composition. Emphasis is placed on machine-learning methods to enable MD-based screening. The resulting framework enables the generation of compound-property databases and the use of advanced statistical modeling to gather insight. The review further summarizes a number of relevant applications.

The Martini Model in Materials Science

“…This introduces a degeneracy in the CG representation, which translates into a reduction of the size of the chemical (compound) space. [ 317 ] Accordingly, this reduction of the size of the chemical space represents also a further speed‐up which can help for screening studies. [ 317 ] Hence, we expect hybrid Martini/machine‐learning schemes to be highly promising in order to efficiently explore the chemical space for different applications.…”

Chemically specific coarse‐graining of polymers: Methods and prospects

“…Last, we remark on the role of ML in computational materials discovery, which remains modest for polymer applications, 35,357,358,372,373 with one impediment being the availability of requisite data. While this limitation can sometimes be overcome by utilizing curated experimental datasets and restricting the design space, 374,375 another emerging strategy, which has been used in diverse applications, is to use CG polymer simulations and surrogate ML predictions to guide the design process.…”

“…379 Because these works were demonstrative in nature, the underlying CG models were not specific to any particular polymer chemistry. Future work would thus benefit from high-throughput methods for defining and parameterizing models 373 (with some viable approaches being previously described). Another option is to use more transferable CG models, such as in the work of Shmilovich et al who used active learning to identify tripeptides with desired aggregation behavior but retained chemical specificity through the Martini model.…”

Featurization Strategies for Polymer Sequence or Composition Design by Machine Learning

“…[5][6][7][8][9] For example, using a limited set of just three different monomer types, there are on the order of 10 47 distinct copolymers that can be generated with degree of polymerization between 10 and 100. Thus, while theory and modeling are invaluable for understanding the origins of observed phenomena and informing the design of specific, well-defined polymer systems, [10][11][12][13][14][15][16][17] intricate studies may severely limit exposure to unknown but promising regions of design space. 18 In addition, resource limitations (time, monetary, or computational) likely preclude exhaustive characterization of combinatorial search spaces.…”

“…28 While flourishing in the domain of "hard" materials and small molecules, applications of ML to polymer design have been relatively limited by comparison for a number of practical and technical reasons. 16,[29][30][31][32][33][34] For example, there are numerous large, open-access databases for small molecules and ordered materials, but data availability and accessibility remains a major challenge for polymer ML. 29,35,36 Presently, this challenge is overcome by either (i) laborious, brute-force data sourcing and curation or (ii) in-house data generation.…”