Oleksii Prykhodko

ORCID: 0000-0002-9694-1192
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Computational Drug Discovery Methods
  • Machine Learning in Materials Science
  • Chemical Synthesis and Analysis
  • Bioinformatics and Genomic Networks
  • Various Chemistry Research Topics
  • Gene expression and cancer classification
  • Protein Structure and Dynamics
  • Genetics, Bioinformatics, and Biomedical Research
  • Process Optimization and Integration

AstraZeneca (Sweden)
2019

Recurrent Neural Networks (RNNs) trained with a set of molecules represented as unique (canonical) SMILES strings, have shown the capacity to create large chemical spaces valid and meaningful structures. Herein we perform an extensive benchmark on models subsets GDB-13 different sizes (1 million, 10,000 1000), variants (canonical, randomized DeepSMILES), two recurrent cell types (LSTM GRU) hyperparameter combinations. To guide benchmarks new metrics were developed that define how well model...

10.1186/s13321-019-0393-0 article EN cc-by Journal of Cheminformatics 2019-11-21

Deep learning methods applied to drug discovery have been used generate novel structures. In this study, we propose a new deep architecture, LatentGAN, which combines an autoencoder and generative adversarial neural network for de novo molecular design. We the method in two scenarios: one random drug-like compounds another target-biased compounds. Our results show that works well both cases. Sampled from trained model can largely occupy same chemical space as training set also substantial...

10.1186/s13321-019-0397-9 article EN cc-by Journal of Cheminformatics 2019-12-01

Recurrent Neural Networks (RNNs) trained with a set of molecules represented as unique (canonical) SMILES strings, have shown the capacity to create large chemical spaces valid and meaningful structures. Herein we perform an extensive benchmark on models subsets GDB-13 different sizes (1 million , 10,000 1,000), variants (canonical, randomized DeepSMILES), two recurrent cell types (LSTM GRU) hyperparameter combinations. To guide benchmarks new metrics were developed that define generated...

10.26434/chemrxiv.8639942.v1 preprint EN 2019-07-05

<p> </p><p>Deep learning methods applied to drug discovery have been used generate novel structures. In this study, we propose a new deep architecture, LatentGAN, which combines an autoencoder and generative adversarial neural network for de novo molecular design. We the method in two scenarios: one random drug-like compounds another target-biased compounds. Our results show that works well both cases: sampled from trained model can largely occupy same chemical space as...

10.26434/chemrxiv.8299544 preprint EN cc-by-nc 2019-06-20

Deep learning methods applied to drug discovery have been used generate novel structures. In this study, we propose a new deep architecture, LatentGAN, which combines an autoencoder and generative adversarial neural network for de novo molecular design. We the method in two scenarios: one random drug-like compounds another target-biased compounds. Our results show that works well both cases: sampled from trained model can largely occupy same chemical space as training set also substantial...

10.26434/chemrxiv.8299544.v4 preprint EN cc-by-nc 2019-09-23

Recurrent Neural Networks (RNNs) trained with a set of molecules represented as unique (canonical) SMILES strings, have shown the capacity to create large chemical spaces valid and meaningful structures. Herein we perform an extensive benchmark on models subsets GDB-13 different sizes (1 million , 10,000 1,000), variants (canonical, randomized DeepSMILES), two recurrent cell types (LSTM GRU) hyperparameter combinations. To guide benchmarks new metrics were developed that define generated...

10.26434/chemrxiv.8639942.v2 preprint EN cc-by 2019-07-30

Deep learning methods applied to drug discovery have been used generate novel structures. In this study, we propose a new deep architecture, LatentGAN, which combines an autoencoder and generative adversarial neural network for de novo molecular design. We the method in two scenarios: one random drug-like compounds another target-biased compounds. Our results show that works well both cases: sampled from trained model can largely occupy same chemical space as training set also substantial...

10.26434/chemrxiv.8299544.v3 preprint EN 2019-09-11

Recently deep learning method has been used for generating novel structures. In the current study, we proposed a new method, LatentGAN, which combine an autoencoder and generative adversarial neural network doing de novo molecule design. We applied structure generation in two scenarios, one is to generate random drug-like compounds other target biased compounds. Our results show that works well both cases, sampled from trained model can largely occupy same chemical space of training set...

10.26434/chemrxiv.8299544.v2 preprint EN cc-by-nc 2019-07-09

Recently deep learning method has been used for generating novel structures. In the current study, we proposed a new method, LatentGAN, which combine an autoencoder and generative adversarial neural network doing de novo molecule design. We applied structure generation in two scenarios, one is to generate random drug-like compounds other target biased compounds. Our results show that works well both cases, sampled from trained model can largely occupy same chemical space of training set...

10.26434/chemrxiv.8299544.v1 preprint EN cc-by-nc 2019-06-20

In recent years, deep learning for de novo molecular generation has become a rapidly growing research area. Recurrent neural networks (RNN) using the SMILES representation is one of most common approaches used. Recent study shows that differentiable computer (DNC) can make considerable improvement over RNN modeling sequential data. current study, DNC been implemented as an extension to REINVENT, RNN-based model already used successfully design. The was benchmarked on its capacity learn...

10.26434/chemrxiv.9758600.v1 preprint EN cc-by-nc 2019-09-04

Recurrent Neural Networks (RNNs) trained with a set of molecules represented as unique (canonical) SMILES strings, have shown the capacity to create large chemical spaces valid and meaningful structures. Herein we perform an extensive benchmark on models subsets GDB-13 different sizes (1 million , 10,000 1,000), variants (canonical, randomized DeepSMILES), two recurrent cell types (LSTM GRU) hyperparameter combinations. To guide benchmarks new metrics were developed that define generated...

10.26434/chemrxiv.8639942 preprint EN 2019-07-05
Coming Soon ...