Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules

Abstract

We report a method to convert discrete representations of molecules to and from a multidimensional continuous representation. This model allows us to generate new molecules for efficient exploration and optimization through open-ended spaces of chemical compounds. A deep neural network was trained on hundreds of thousands of existing chemical structures to construct three coupled functions: an encoder, a decoder, and a predictor. The encoder converts the discrete representation of a molecule into a real-valued continuous vector, and the decoder converts these continuous vectors back to discrete molecular representations. The predictor estimates chemical properties from the latent continuous vector representation of the molecule. Continuous representations of molecules allow us to automatically generate novel chemical structures by performing simple operations in the latent space, such as decoding random vectors, perturbing known chemical structures, or interpolating between molecules. Continuous representations also allow the use of powerful gradient-based optimization to efficiently guide the search for optimized functional compounds. We demonstrate our method in the domain of drug-like molecules and also in a set of molecules with fewer that nine heavy atoms.

Deep Genomic Signature for early metastasis prediction in prostate cancer

“…Deep Patient, an unsupervised deep feature learning method based on Stacked DAE, was developed to predict future of patients with different cancers from Electronic Health Records data (Miotto et al, 2016). VAE and AAE were also successfully utilized in designing new molecules with desired properties for drug discovery purposes (Gómez-Bombarelli et al, 2016;Kadurin et al, 2017). Moreover, VAE was also able to capture patterns in the gene expression pan-cancer data for specific tissues (Way and Greene, 2018).…”

Exploring differential evolution for inverse QSAR analysis

“…For this purpose, the use of Gaussian mixture models and cluster-wise MLR left considerable room for improvement, due to its multi-parametric nature and tendency of overfitting if training data were organized into large number of clusters 14 . Recently, autoencoder modeling was proposed as an approach for two-stage inverse QSAR 16 . Continuous latent space, corresponding to a descriptor space, is constructed on the basis of encoding a line notation of a molecule by recurrent neural networks (RNNs).…”

“…As such, the approach does not depend on chosen descriptors and has the potential to automatically address two-stage inverse QSAR in a single step. However, the generation of new valid line notations (SMILES strings) for chemical structures corresponding to optimized coordinates was difficult in a case study designing organic light-emitting diodes 16 .…”

“…al. 1 who use generative models to approach the inverse-QSAR problem. Their work concerns building and optimizing QSAR equations in the latent space of an autoencoder and subsequently decoding optimized points into molecular structures.…”

Prediction of pKa values for drug-like molecules using semiempirical quantum chemical methods

“…For example, Settimo et al (2013) have recently shown that the empirical methods can fail for some amines, which represent a large fraction of drugs currently on the market or in development. This problem could make these methods difficult to apply to computational exploration of chemical space (Rupakheti et al, 2016;Gómez-Bombarelli et al, 2016) where molecules with completely new chemical substructures are likely to be encountered.…”