Enriched optimization of molecular properties under constraints: an electrochromic example

Rinderspacher, B. Christopher; Elward, Jennifer M.

doi:10.1039/c7me00126f

Cited by 4 publications

(11 citation statements)

References 94 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…R diversity mismatch = − r generated diversity − r training diversity (28) Another diversity measure that has been employed is uniqueness, which aims to reduce the number of repeat molecules. The uniqueness reward R uniqueness ∈ (0, 1] is defined as:…”

Section: Diversity and Noveltymentioning

confidence: 99%

“…While much of the work so far has focused on deep generative modeling for drug molecules, 24 there are many other application domains which are benefiting from the application of deep learning to lead generation and screening, such as organic light emitting diodes, 25 organic solar cells, 26 energetic materials, 10,27 electrochromic devices, 28 polymers, 29 polypeptides, [30][31][32] and metal organic frameworks. 33,34 Our review touches on four major issues we have observed in the field.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Deep learning for molecular design—a review of the state of the art

Elton

Boukouvalas

Fuge

et al. 2019

Mol. Syst. Des. Eng.

556

516

View full text Add to dashboard Cite

In the space of only a few years, deep generative modeling has revolutionized how we think of artificial creativity, yielding autonomous systems which produce original images, music, and text. Inspired by these successes, researchers are now applying deep generative modeling techniques to the generation and optimization of molecules-in our review we found 45 papers on the subject published in the past two years. These works point to a future where such systems will be used to generate lead molecules, greatly reducing resources spent downstream synthesizing and characterizing bad leads in the lab. In this review we survey the increasingly complex landscape of models and representation schemes that have been proposed. The four classes of techniques we describe are recursive neural networks, autoencoders, generative adversarial networks, and reinforcement learning. After first discussing some of the mathematical fundamentals of each technique, we draw high level connections and comparisons with other techniques and expose the pros and cons of each. Several important high level themes emerge as a result of this work, including the shift away from the SMILES string representation of molecules towards more sophisticated representations such as graph grammars and 3D representations, the importance of reward function design, the need for better standards for benchmarking and testing, and the benefits of adversarial training and reinforcement learning over maximum likelihood based training.

show abstract

Section: Diversity and Noveltymentioning

confidence: 99%

mentioning

confidence: 99%

Deep learning for molecular design—a review of the state of the art

Elton

Boukouvalas

Fuge

et al. 2019

Mol. Syst. Des. Eng.

556

516

View full text Add to dashboard Cite

show abstract

“…Consequently, a number of methods have been developed to explore chemical subspaces efficiently. ,,,− Lack of structure may express itself in property roughness or constraint-induced feasibility islands. Prominent approaches to address these as well as search space discreteness are stochastic methods, ,,, in particular genetic algorithms (GAs). ,, GAs address the problem of property roughness by using populations of candidates of which not all members are solely evolved by performance.…”

Section: Introductionmentioning

confidence: 99%

Heuristic Global Optimization in Chemical Compound Space

Rinderspacher

2020

J. Phys. Chem. A

Self Cite

View full text Add to dashboard Cite

We report developments in combinatorial optimization under constraints in chemical space. We considered random functions, which serve as a baseline to measure performance, and constrained optimizations over two databases of electrochromic molecules (∼5000 and 10 12 , respectively). These problems were optimized using sequential heuristic next-neighbor search (HNNS) (introduced in Elward and Rinderspacher Phys. Chem. Chem. Phys. 2015, 17, 24322 and Elward and Rinderspacher Mol. Syst. Des. Eng. 2018, 3, 485) and kernel-based efficient global optimization (EGO) with two reordering strategies. In addition to the average ordering method introduced with sequential HNNS, a new reordering based on a locally separable linear model is formulated and applied. Presented is the analysis of periodic kernels for EGO: the modified Dirichlet kernels D ̃n(x) = ∑ i n [cos(2πix) − (−1) i ]/i, the minimalist kernel K N (x) = cos(πx/N) 2 /2 − δ x /2, and an analogue to the popular Gaussian kernel, G N σ (x) = e −|x mod N| 2 /(2σ 2 ), where n is the order of the Dirichlet kernel, N is the number of choices on a combinatorial site, and σ is a broadening factor. We find that reordering is pivotal for high optimization efficiency, in both the average and worst-case scenarios. The new linear-estimation ordering paradigm is superior in the chemistry context. Furthermore, with judicious use of hyperparameters and algorithmic choices, EGO outperforms HNNS. The global optimum for the small chemical problem is found with >98% likelihood for all global search methods employing a linear reordering heuristic for organizing the search space.

show abstract

“…For logical analysis, we propose use of a feedforward deep neural network model (multilayer perceptron). , The backpropagation algorithm does the training of such a machine learning model . Different computer and machine learning methods are able to solve a lot of biological and chemical problems, in particular, establishing the structure of proteins and their interactions, as well as predicting the effect of mutations on their assembly. , …”

mentioning

confidence: 99%

“…25 Different computer and machine learning methods are able to solve a lot of biological and chemical problems, in particular, establishing the structure of proteins and their interactions, as well as predicting the effect of mutations on their assembly. 26,27 Multilayer perceptron supervised learning allows the detection of individual states on an eGaIn/hydrogel interface that depends on hydrogel composition. A database is necessary for model training.…”

mentioning

confidence: 99%

Tick-Borne Encephalitis Electrochemical Detection by Multilayer Perceptron on Liquid–Metal Interface

Ivanov

Nikolaev

Stekolshchikova

et al. 2020

ACS Appl. Bio Mater.

View full text Add to dashboard Cite

This work depicts an electrochemical hydrogel–eutectic gallium indium alloy interface for the detection of tick-borne encephalitis (TBE) virus. This interface allows recording of nonlinear current–voltage responses, depending on the composition of the hydrogel. The current–voltage data for the machine learning model are trained by a multilayer perceptron. This model accurately recognizes the TBE antibody, antigen, and an antibody–antigen complex in mixture with interfering bovine serum albumin with 93% accuracy. Thus, this interface can be used as a convenient method for expressed viruses and pathogens detection.

show abstract

Enriched optimization of molecular properties under constraints: an electrochromic example

Abstract: We present a deterministic optimization procedure of molecular properties that ensures diverse coverage of the given chemical compound search space.

Cited by 4 publications

References 94 publications

Deep learning for molecular design—a review of the state of the art

Deep learning for molecular design—a review of the state of the art

Heuristic Global Optimization in Chemical Compound Space

Tick-Borne Encephalitis Electrochemical Detection by Multilayer Perceptron on Liquid–Metal Interface

Contact Info

Product

Resources

About