CReM: chemically reasonable mutations framework for structure generation

Polishchuk, Pavel

doi:10.21203/rs.2.23402/v1

“…The results can be found in Table 3. In addition to the GuacaMol baseline models, we compare ourselves to two recent methods, namely CReM [5] and MSO [9]. The three first columns for EvoMol correspond to the mean scores obtained on 10 runs for each initial conditions and parameters.…”

Section: Case 2: Guacamolmentioning

confidence: 99%

EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation

Leguy

¹

,

Cauchy

²

,

Duval

³

et al. 2020

Preprint

View full text Add to dashboard Cite

The objective of this work is to design a molecular generator capable of exploring known as well as unfamiliar areas of the chemical space. Our method must be flexible to adapt to very different problems. Therefore, it has to be able to work with or without the influence of prior data and knowledge. Moreover, regardless of the success, it should be as interpretable as possible to allow for diagnosis and improvement. We propose here a new open source generation method using an evolutionary algorithm to sequentially build molecular graphs. It is independent of starting data and can generate totally unseen compounds. To be able to search a large part of the chemical space, we define an original set of 7 generic mutations close to the atomic level. Our method achieves excellent performances and even records on the QED, penalised logP, SAscore, CLscore as well as the set of goal-directed functions defined in GuacaMol. To demonstrate its flexibility, we tackle a very different objective issued from the organic molecular materials domain. We show that EvoMol can generate sets of optimised molecules having high energy HOMO or low energy LUMO, starting only from methane. We can also set constraints on a synthesizability score and structural features. Finally, the interpretability of EvoMol allows for the visualisation of its exploration process as a chemically relevant tree.

show abstract

“…The results can be found in Table 3 . In addition to the GuacaMol baseline models, Graph GA [ 27 ] and SMILES LSTM [ 48 ], we compare ourselves to two recent methods, namely CReM [ 5 ] and MSO [ 9 ]. The three first columns for EvoMol correspond to the mean scores obtained on 10 runs for each initial conditions and parameters.…”

Section: Resultsmentioning

confidence: 99%

“…To limit the number of steps and to improve the likeliness of the solutions, they were commonly based on the combination of fragments rather than mutating the molecules at atomic level. The interest in evolutionary algorithms has decreased with the emergence of deep learning for molecular generation, although very recently, a new and efficient fragment based method was designed [ 5 ]. In the mid 2010s, Aspuru-Guzik and coll.…”

Section: Introductionmentioning

confidence: 99%

EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation

Leguy¹,

Cauchy

²

,

Glavatskikh

³

et al. 2020

View full text Add to dashboard Cite

The objective of this work is to design a molecular generator capable of exploring known as well as unfamiliar areas of the chemical space. Our method must be flexible to adapt to very different problems. Therefore, it has to be able to work with or without the influence of prior data and knowledge. Moreover, regardless of the success, it should be as interpretable as possible to allow for diagnosis and improvement. We propose here a new open source generation method using an evolutionary algorithm to sequentially build molecular graphs. It is independent of starting data and can generate totally unseen compounds. To be able to search a large part of the chemical space, we define an original set of 7 generic mutations close to the atomic level. Our method achieves excellent performances and even records on the QED, penalised logP, SAscore, CLscore as well as the set of goal-directed functions defined in GuacaMol. To demonstrate its flexibility, we tackle a very different objective issued from the organic molecular materials domain. We show that EvoMol can generate sets of optimised molecules having high energy HOMO or low energy LUMO, starting only from methane. We can also set constraints on a synthesizability score and structural features. Finally, the interpretability of EvoMol allows for the visualisation of its exploration process as a chemically relevant tree.

show abstract

“…GANs have also been used for creating latent space and inverse QSAR that was searched to generate novel compounds with predicted activity against the dopamine receptor type 2 74 . Fragment‐based structure generation has been proposed through knowledge and use of synthetic complexity, chemically valid structures, novelty, and diversity of compounds 75 . It has been used in multiparameter optimization in de novo design based on an actor/critic model both using bidirectional long short‐term memory (LSTM) networks, 76 where the AI method learns how to generate new compounds with desired properties from an initial set of lead molecules and improving them by replacing their fragments.Thus obtained were 93% of the generated molecules as chemically valid and over 33% satisfied the targeted objectives, while there were none in the initial set 76 …”

Section: Applications Of ML In Drug Designmentioning

confidence: 99%

Machine learning, artificial intelligence, and data science breaking into drug design and neglected diseases

Peña-Guerrero

¹

,

Nguewa

²

,

García-Sosa

³

2021

WIREs Comput Mol Sci

View full text Add to dashboard Cite

Machine learning (ML) is becoming capable of transforming biomolecular interaction description and calculation, promising an impact on molecular and drug design, chemical biology, toxicology, among others. The first improvements can be seen from biomolecule structure prediction to chemical synthesis, molecular generation, mechanism of action elucidation, inverse design, polypharmacology, organ or issue targeting of compounds, property and multiobjective optimization. Chemical design proposals from an algorithm may be inventive and feasible. Challenges remain, with the availability, diversity, and quality of data being critical for developing useful ML models; marginal improvement seen in some cases, as well as in the interpretability, validation, and reuse of models. The ultimate aim of ML should be to facilitate options for the scientist to propose and undertake ideas and for these to proceed faster. Applications are ripe for transformative results in understudied, neglected, and rare diseases, where new data and therapies are strongly required. Progress and outlook on these themes are provided in this study. This article is categorized under: Structure and Mechanism > Computational Biochemistry and Biophysics Structure and Mechanism > Molecular Structures

show abstract

CReM: chemically reasonable mutations framework for structure generation

Cited by 14 publications

References 28 publications

EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation

EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation

EvoMol: a flexible and interpretable evolutionary algorithm for unbiased de novo molecular generation

Machine learning, artificial intelligence, and data science breaking into drug design and neglected diseases

Contact Info

Product

Resources

About