Generative Models for Chemical Structures

Abstract

We apply recently developed techniques for pattern recognition to construct a generative model for chemical structure. This approach can be viewed as ligand-based de novo design. We construct a statistical model describing the structural variations present in a set of molecules which may be sampled to generate new structurally similar examples. We prevent the possibility of generating chemically invalid molecules, according to our implicit hydrogen model, by projecting samples onto the nearest chemically valid molecule. By populating the input set with molecules that are active against a target, we show how new molecules may be generated that will likely also be active against the target.

Granular computing, computational intelligence, and the analysis of non-geometric input spaces

“…However, many important applications can be tackled by representing patterns as ''non-geometric'' entities. For instance, it is possible to cite applications in document analysis (Bunke and Riesen 2011), solubility of E. coli proteome , bio-molecules recognition (Ceroni et al 2007;Rupp and Schneider 2010), chemical structures generation (White and Wilson 2010), image analysis (Serratosa et al 2013;Morales-González et al 2014), and scene understanding (Brun et al 2014). The availability of interesting datasets containing non-geometric data motivated the development of pattern recognition and soft computing techniques on such domains Rossi et al 2015;Fischer et al 2015;Lange et al 2015;Schleif 2014;Bianchi et al 2015).…”

Inverse‐QSPR for de novo Design: A Review

“…White and Wilson used DRM within a graph‐based generation algorithm. Authors started to compute fragments from a set of molecules and to extract their distribution within the set.…”

Designing Anticancer Peptides by Constructive Machine Learning

“…Constructive artificial intelligence (AI) has recently received increased attention for de novo molecule design . When applied to drug design, this concept allows the computational generation of novel chemical structures without having to explicitly specify building block libraries, chemical reaction queries, descriptor‐based representations and/or rules for molecule assembly.…”

“…When applied to drug design, this concept allows the computational generation of novel chemical structures without having to explicitly specify building block libraries, chemical reaction queries, descriptor‐based representations and/or rules for molecule assembly. While several studies have dealt with constructive models in chemistry from a theoretical point of view, the first constructive AI model has only recently been experimentally validated by generating novel nuclear receptor modulators . Recurrent neural networks (RNNs) bear particular promise as de novo design tools .…”

“…Six of the active peptides killedM CF7 cancer cells without affecting human erythrocytes with at least threefold selectivity.T hese results advocate constructivem achine learning for the automated design of peptides with desired biological activities.Constructive artificial intelligence (AI) has recently received increaseda ttention for de novo molecule design. [1][2][3][4][5] When applied to drug design,t his concept allows the computational generation of novel chemical structures withouth aving to explicitly specify buildingb lock libraries, chemical reactionq ueries, descriptor-based representations and/or rules for molecule assembly.W hile severals tudies have dealt with constructive modelsi nc hemistry from at heoretical point of view, [1][2][3][4] the first constructive AI model haso nly recently been experimentally validated by generating novel nuclear receptor modulators. [5] Recurrent neural networks (RNNs) bear particularp romise as de novo design tools.…”