Comparison of structure- and ligand-based scoring functions for deep generative models: a GPCR case study

Deep generative models have shown the ability to devise both valid and novel chemistry, which could significantly accelerate the identification of bioactive compounds. Many current models, however, use molecular descriptors or ligand-based predictive methods to guide molecule generation towards a desirable property space. This restricts their application to relatively data-rich targets, neglecting those where little data is available to sufficiently train a predictor. Moreover, ligand-based models often bias molecule generation towards previously established chemical space, thereby limiting their ability to identify truly novel chemotypes. In this work, we assess the ability of using molecular docking via Glide – a structure-based approach – as a scoring function to guide the deep generative model REINVENT and compare model performance and behaviour to a ligand-based scoring function. Additionally, we modify the previously published MOSES benchmarking dataset to remove any induced bias towards non-protonatable groups. We also propose a new metric to measure dataset diversity, which is less confounded by the distribution of heavy atom count than the commonly used internal diversity metric. With respect to the main findings, we found that when optimizing the docking score against DRD2, the model improves predicted ligand affinity beyond that of known DRD2 active molecules. In addition, generated molecules occupy complementary chemical and physicochemical space compared to the ligand-based approach, and novel physicochemical space compared to known DRD2 active molecules. Furthermore, the structure-based approach learns to generate molecules that satisfy crucial residue interactions, which is information only available when taking protein structure into account. Overall, this work demonstrates the advantage of using molecular docking to guide de novo molecule generation over ligand-based predictors with respect to predicted affinity, novelty, and the ability to identify key interactions between ligand and protein target. Practically, this approach has applications in early hit generation campaigns to enrich a virtual library towards a particular target, and also in novelty-focused projects, where de novo molecule generation either has no prior ligand knowledge available or should not be biased by it.

Download Full-text

Comparison of Structure- and Ligand-Based Scoring Functions for Deep Generative Models: A GPCR Case Study

10.26434/chemrxiv.14138147 ◽

2021 ◽

Author(s):

Morgan Thomas ◽

Rob Smith ◽

Noel M. O’Boyle ◽

Chris de Graaf ◽

Andreas Bender

Keyword(s):

Molecular Docking ◽

De Novo ◽

Chemical Space ◽

Heavy Atom ◽

Model Performance ◽

Scoring Function ◽

Generative Models ◽

Scoring Functions ◽

Property Space

Deep generative models have shown the ability to devise both valid and novel chemistry, which could significantly accelerate the identification of bioactive compounds. Many current models, however, use molecular descriptors or ligand-based predictive methods to guide molecule generation towards a desirable property space. This restricts their application to relatively data-rich targets, neglecting those where little data is available to sufficiently train a predictor. Moreover, ligand-based models often bias molecule generation towards previously established chemical space, thereby limiting their ability to identify truly novel chemotypes. In this work, we assess the ability of using molecular docking via Glide – a structure-based approach – as a scoring function to guide the deep generative model REINVENT and compare model performance and behaviour to a ligand-based scoring function. Additionally, we modify the previously published MOSES benchmarking dataset to remove any induced bias towards non-protonatable groups. We also propose a new metric to measure dataset diversity, which is less confounded by the distribution of heavy atom count than the commonly used internal diversity metric. With respect to the main findings, we found that when optimizing the docking score against DRD2, the model improves predicted ligand affinity beyond that of known DRD2 active molecules. In addition, generated molecules occupy complementary chemical and physicochemical space compared to the ligand-based approach, and novel physicochemical space compared to known DRD2 active molecules. Furthermore, the structure-based approach learns to generate molecules that satisfy crucial residue interactions, which is information only available when taking protein structure into account. Overall, this work demonstrates the advantage of using molecular docking to guide de novo molecule generation over ligand-based predictors with respect to predicted affinity, novelty, and the ability to identify key interactions between ligand and protein target. Practically, this approach has applications in early hit generation campaigns to enrich a virtual library towards a particular target, and also in novelty-focused projects, where de novo molecule generation either has no prior ligand knowledge available or should not be biased by it.

Download Full-text

GEN: Highly Efficient SMILES Explorer Using Autodidactic Generative Examination Networks

10.26434/chemrxiv.9796874.v1 ◽

2019 ◽

Cited By ~ 1

Author(s):

Ruud van Deursen ◽

Peter Ertl ◽

Igor Tetko ◽

Guillaume Godin

Keyword(s):

De Novo ◽

Chemical Space ◽

Quality Criteria ◽

Generative Models ◽

Target Space ◽

Statistical Quality Control ◽

Stable Model ◽

Online Examination ◽

Property Space

Recurrent neural networks have been widely used to generate millions of de novo molecules in a known chemical space. These deep generative models are typically setup with LSTM or GRU units and trained with canonical SMILES. In this study, we introduce a new robust architecture, Generative Examination Network GEN, based on bidirectional RNNs with concatenated sub-models to learn and generate molecular SMILES within a trained target space. GENs autonomously learn the target space in a few epochs while being subjected to an independent online examination to measure the quality of the generated set. Here we have used online statistical quality control (SQC) on the percentage of valid molecular SMILES as examination measure to select the earliest available stable model weights. Very high levels of valid SMILES (95-98%) can be generated using multiple parallel encoding layers in combination with SMILES augmentation using unrestricted SMILES randomization. Our architecture combines an excellent novelty rate (85-90%) while generating SMILES with strong conservation of the property space (95-99%). Our flexible examination mechanism is open to other quality criteria.

Download Full-text

GEN: Highly Efficient SMILES Explorer Using Autodidactic Generative Examination Networks

10.26434/chemrxiv.9796874 ◽

2019 ◽

Author(s):

Ruud van Deursen ◽

Peter Ertl ◽

Igor Tetko ◽

Guillaume Godin

Keyword(s):

De Novo ◽

Chemical Space ◽

Quality Criteria ◽

Generative Models ◽

Target Space ◽

Statistical Quality Control ◽

Stable Model ◽

Online Examination ◽

Property Space

Recurrent neural networks have been widely used to generate millions of de novo molecules in a known chemical space. These deep generative models are typically setup with LSTM or GRU units and trained with canonical SMILES. In this study, we introduce a new robust architecture, Generative Examination Network GEN, based on bidirectional RNNs with concatenated sub-models to learn and generate molecular SMILES within a trained target space. GENs autonomously learn the target space in a few epochs while being subjected to an independent online examination to measure the quality of the generated set. Here we have used online statistical quality control (SQC) on the percentage of valid molecular SMILES as examination measure to select the earliest available stable model weights. Very high levels of valid SMILES (95-98%) can be generated using multiple parallel encoding layers in combination with SMILES augmentation using unrestricted SMILES randomization. Our architecture combines an excellent novelty rate (85-90%) while generating SMILES with strong conservation of the property space (95-99%). Our flexible examination mechanism is open to other quality criteria.

Download Full-text

SMILES-Based Deep Generative Scaffold Decorator for De-Novo Drug Design

10.26434/chemrxiv.11638383.v1 ◽

2020 ◽

Author(s):

Josep Arús-Pous ◽

Atanas Patronov ◽

Esben Jannik Bjerrum ◽

Christian Tyrchan ◽

Jean-Louis Reymond ◽

...

Keyword(s):

Data Augmentation ◽

De Novo ◽

Chemical Space ◽

Generative Models ◽

Synthetic Chemistry ◽

Specific Knowledge ◽

De Novo Drug Design ◽

Wide Range ◽

Small Set ◽

Synthetic Approaches

Molecular generative models trained with small sets of molecules represented as SMILES strings are able to generate large regions of the chemical space. Unfortunately, due to the sequential nature of SMILES strings, these models are not able to generate molecules given a scaffold (i.e. partially-built molecules with explicit attachment points). Herein we report a new SMILES-based molecular generative architecture that generates molecules from scaffolds and can be trained from any arbitrary molecular set. This is possible thanks to a new molecular set pre-processing algorithm that exhaustively cuts all combinations of acyclic bonds of every molecule, obtaining a large number of scaffold-decorations combinations. Moreover, it serves as a data augmentation technique and can be readily coupled with randomized SMILES to obtain even better results with small sets. Two examples showcasing the potential of the architecture in medicinal and synthetic chemistry are described: First, models were trained with a training set obtained from a small set of Dopamine Receptor D2 (DRD2) active modulators and were able to meaningfully decorate a wide range of scaffolds and obtain molecular series predicted active on DRD2. Second, a larger set of drug-like molecules from ChEMBL was selectively sliced using synthetic chemistry constraints (RECAP rules). Moreover, the resulting scaffold-decorations were filtered to only allow decorations that were fragment-like. This allowed models trained with this dataset to selectively decorate diverse scaffolds with fragments that were generally predicted to be synthesizable and attachable to the scaffold using known synthetic approaches. In both cases, the models were already able to decorate molecules using specific knowledge without the need to add it with other techniques, such as reinforcement learning. We envision that this architecture will become a useful addition to the already existent architectures for de-novo molecular generation.

Download Full-text

Memory-assisted reinforcement learning for diverse molecular de novo design

10.21203/rs.3.rs-52871/v2 ◽

2020 ◽

Author(s):

Thomas Blaschke ◽

Ola Engkvist ◽

Jürgen Bajorath ◽

Hongming Chen

Keyword(s):

Reinforcement Learning ◽

De Novo ◽

Molecular Design ◽

Chemical Space ◽

Scoring Function ◽

Chemical Diversity ◽

Chemical Structures ◽

Machine Learning Model ◽

Low Diversity ◽

De Novo Molecular Design

Abstract In de novo molecular design, recurrent neural networks (RNN) have been shown to be effective methods for sampling and generating novel chemical structures. Using a technique called reinforcement learning (RL), an RNN can be tuned to target a particular section of chemical space with optimized desirable properties using a scoring function. However, ligands generated by current RL methods so far tend to have relatively low diversity, and sometimes even result in duplicate structures when optimizing towards desired properties. Here, we propose a new method to address the low diversity issue in RL for molecular design. Memory-assisted RL is an extension of the known RL, with the introduction of a so-called memory unit. As proof of concept, we applied our method to generate structures with a desired AlogP value. In a second case study, we applied our method to design ligands for the dopamine type 2 receptor and the 5-hydroxytryptamine type 1A receptor. For both receptors, a machine learning model was developed to predict whether generated molecules were active or not for the receptor. In both case studies, it was found that memory-assisted RL led to the generation of more compounds predicted to be active having higher chemical diversity, thus achieving better coverage of chemical space of known ligands compared to established RL methods.

Download Full-text

Prediction of Compound Synthesis Accessibility Based on Reaction Knowledge Graph

10.26434/chemrxiv.14608296.v1 ◽

2021 ◽

Author(s):

Baiqing Li ◽

Hongming Chen

Keyword(s):

Deep Learning ◽

Prediction Models ◽

De Novo ◽

Quantitative Estimation ◽

Reaction Network ◽

Generative Models ◽

Machine Learning Algorithms ◽

Scoring Functions ◽

Synthetic Accessibility ◽

Learning Machine

<a>With the increasing application of deep learning based generative models for de novo molecule design, quantitative estimation of molecular synthetic accessibility becomes a crucial factor for prioritizing the structures generated from generative models. On the other hand, it is also useful for helping prioritization of hit/lead compounds and guiding retro-synthesis analysis. In current study, based on the USPTO and Pistachio reaction datasets, we created a chemical reaction network, in which a depth-first search was performed for identification of the reaction paths of product compounds. This reaction dataset was then used to build predictive model for distinguishing the organic compounds either as easy synthesize (ES) or hard-to synthesize (HS) classes. Three synthesis accessibility (SA) models were built using deep learning/machine learning algorithms. The comparison between our three SA scoring functions with other existing synthesis accessibility scoring schemes, such as SYBA, SCScore, SAScore were also carried out. and the graph based deep learning model outperforms those existing SA scores. Our results show that prediction models based on historical reaction knowledge could be a useful tool for measuring molecule complexity and estimating molecule SA.</a>

Download Full-text

METADOCK 2: a high-throughput parallel metaheuristic scheme for molecular docking

Bioinformatics ◽

10.1093/bioinformatics/btz958 ◽

2020 ◽

Author(s):

Baldomero Imbernón ◽

Antonio Serrano ◽

Andrés Bueno-Crespo ◽

José L Abellán ◽

Horacio Pérez-Sánchez ◽

...

Keyword(s):

Molecular Docking ◽

Graphics Processing Units ◽

Degrees Of Freedom ◽

Scoring Function ◽

Optimization Procedure ◽

Supplementary Information ◽

Scoring Functions ◽

Autodock Vina ◽

Molecular Conformations ◽

Computational Performance

Abstract Motivation Molecular docking methods are extensively used to predict the interaction between protein–ligand systems in terms of structure and binding affinity, through the optimization of a physics-based scoring function. However, the computational requirements of these simulations grow exponentially with: (i) the global optimization procedure, (ii) the number and degrees of freedom of molecular conformations generated and (iii) the mathematical complexity of the scoring function. Results In this work, we introduce a novel molecular docking method named METADOCK 2, which incorporates several novel features, such as (i) a ligand-dependent blind docking approach that exhaustively scans the whole protein surface to detect novel allosteric sites, (ii) an optimization method to enable the use of a wide branch of metaheuristics and (iii) a heterogeneous implementation based on multicore CPUs and multiple graphics processing units. Two representative scoring functions implemented in METADOCK 2 are extensively evaluated in terms of computational performance and accuracy using several benchmarks (such as the well-known DUD) against AutoDock 4.2 and AutoDock Vina. Results place METADOCK 2 as an efficient and accurate docking methodology able to deal with complex systems where computational demands are staggering and which outperforms both AutoDock Vina and AutoDock 4. Availability and implementation https://[email protected]/Baldoimbernon/metadock_2.git. Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

Applications of Autoencoders along with Deep Learning Techniques to generate valid molecules

Journal of Physics Conference Series ◽

10.1088/1742-6596/2070/1/012125 ◽

2021 ◽

Vol 2070 (1) ◽

pp. 012125

Author(s):

T Sesha Sai Aparna ◽

T Anuradha

Keyword(s):

Deep Learning ◽

Drug Design ◽

De Novo ◽

Chemical Space ◽

Solution Space ◽

Threshold Value ◽

Activation Function ◽

Generative Models ◽

Pharmaceutical Chemical ◽

De Novo Drug Design

Abstract From the moment of identifying the fundamental cause of an illness to its availability in the marketplace, it takes an average of 10 years and almost $2.6 billion dollars to develop a medication. We’re actually hunting for a needle in a haystack, which takes a lot of time, effort, and money. In a solution space of between 1030 and 10100 synthetically viable compounds, we’re seeking for the one molecule that can turn off a disease at the molecular level. The chemical solution space is just too large to adequately screen for the desired molecule. Only a small percentage of the synthetically viable compounds for wet lab research are stored in pharmaceutical chemical repositories. Computational de novo drug design can be used to explore this vast chemical space and develop previously undesigned compounds. Computational drug design can cut the amount of time spent in the discovery phase in half, resulting in a shorter time to market and lower drug prices. Deep learning and artificial intelligence (AI) have opened up new perspectives in cheminformatics, especially in molecules generative models. Recurrent neural networks (RNNs) trained with molecules in the SMILES text format, in particular, are very good at exploring the chemical space. Two baseline models were created for generating molecules, one of the model includes an encoder that takes SMILES as input and then develops a deep generative LSTM model which acts as a hidden layer and the output from layers acts as an input to the decoder. The other baseline model acts the same as the above-mentioned model but it includes latent space, it is simply a representation of compressed data that bring related data points closer together physically. To learn data properties and find simpler data representations for analysis, and weights which are obtained from the previous model to generate more efficient molecules. Then created a custom function to play with the temperature of the softmax activation function which creates a threshold value for the valid molecules to generate. This model enables us to produce new molecules through successful exploration.

Download Full-text

Comparative Study of Deep Generative Models on Chemical Space Coverage (v18)

10.26434/chemrxiv.13234289.v3 ◽

2021 ◽

Author(s):

Jie Zhang ◽

Rocío Mercado ◽

Ola Engkvist ◽

Hongming Chen

Keyword(s):

Deep Learning ◽

Functional Groups ◽

De Novo ◽

Molecular Design ◽

Chemical Space ◽

Generative Models ◽

Molecular Structures ◽

Reference Database ◽

Ring Systems ◽

Adversarial Networks

In recent years, deep molecular generative models have emerged as novel methods for de novo molecular design. Thanks to the rapid advance of deep learning techniques, deep learning architectures such as recurrent neural networks, generative autoencoders, and adversarial networks, to give a few examples, have been employed for constructing generative models. However, so far the metrics used to evaluate these deep generative models are not discriminative enough to separate the performance of various state-of-the-art generative models. This work presents a novel metric for evaluating deep molecular generative models; this new metric is based on the chemical space coverage of a reference database, and compares not only the molecular structures, but also the ring systems and functional groups, reproduced from a reference dataset of 1M structures. In this study, the performance of 7 different molecular generative models was compared by calculating their structure and substructure coverage of the GDB-13 database while using a 1M subset of GDB-13 for training. Our study shows that the performance of various generative models varies significantly using the benchmarking metrics introduced herein, such that generalization capability of the generative model can be clearly differentiated. Additionally, the coverage of ring systems and functional groups existing in GDB-13 was also compared between the models. Our study provides a useful new metric that can be used for evaluating and comparing generative models.

Download Full-text