Deep Learning in Drug Discovery and Medicine; Scratching the Surface

Dana, Dibyendu; Gadhiya, Satishkumar; Surin, Luce G St; Li, David; Naaz, Farha; Ali, Quaisar; Paka, Latha; Yamin, Michael; Narayan, Mahesh; Goldberg, Itzhak D.; Narayan, Prakash

doi:10.3390/molecules23092384

Cited by 91 publications

(45 citation statements)

References 67 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, a more data‐hungry ML algorithm, deep learning (DL), which has gained great success in a wide variety of applications, such as computer vision, speech recognition, computer games, and natural language processing, has also attracted considerable interest from computational chemists and medicinal chemists. Up to now, various reviews related to the applications of ML or DL in drug design and discovery have been published . Ain et al and Khamis et al summarized the advances of ML‐based SFs before 2015 in two comprehensive reviews about protein–ligand binding affinity prediction and SBVS, but DL has just begun to rise in the field of drug discovery in 2015 .…”

Section: Introductionmentioning

confidence: 99%

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

Shen

Ding

Wang

et al. 2019

WIREs Comput Mol Sci

197

180

View full text Add to dashboard Cite

Molecule docking has been regarded as a routine tool for drug discovery, but its accuracy highly depends on the reliability of scoring functions (SFs). With the rapid development of machine learning (ML) techniques, ML-based SFs have gradually emerged as a promising alternative for protein-ligand binding affinity prediction and virtual screening, and most of them have shown significantly better performance than a wide range of classical SFs. Emergence of more data-hungry deep learning (DL) approaches in recent years further fascinates the exploitation of more accurate SFs. Here, we summarize the progress of traditional ML-based SFs in the last few years and provide insights into recently developed DL-based SFs. We believe that the continuous improvement in ML-based SFs can surely guide the early-stage drug design and accelerate the discovery of new drugs. This article is categorized under:Computer and Information Science > Chemoinformaticsdeep learning, machine learning, molecular docking, scoring function, structure-based drug design | INTRODUCTIONTraditional drug discovery largely relies on the application of high-throughput screening, an experimental technique with acceptable performance but high cost and low efficiency. 1 With the rapid development of computational chemistry and computer technology, computer-aided drug design (CADD) has gradually emerged as a powerful technique in the design and development of new drug candidates in the past three decades. 2 Virtual screening (VS), an important branch of CADD, can enrich potential actives from large virtual compound libraries through in silico methods rather than real experiments, which can not only accelerate the process of drug discovery but also greatly reduce the time and resource cost. 3-5 Depending on whether the three-dimensional (3D) structure of a target is used or not, VS approaches can be classified into two major categories: ligand-based virtual screening (LBVS) and structure-based virtual screening (SBVS). 6 LBVS aims to discover active molecules through the models developed based on a set of known ligands of a target of interest, which may limit its capability to find novel chemotypes. Compared with LBVS, SBVS is considered to be a better choice to discover novel active compounds if the 3D structure of a given target is available. 7 Chao Shen and Junjie Ding are equivalent first authors.

show abstract

Section: Introductionmentioning

confidence: 99%

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

Shen

Ding

Wang

et al. 2019

WIREs Comput Mol Sci

197

180

View full text Add to dashboard Cite

show abstract

“…Secondly, the current deep neural networks designed for the ligand-based models also use some molecular fingerprints, like ECFP [34], as their input data. This type of input encoding restricts the feature discovery to the composition of the specific molecular structures that are defined by the fingerprinting procedure [29], [35], which eliminates its capacity to discover the arbitrary features. Thirdly, as these models are blind towards the target, they cannot elucidate the potential molecular interactions.…”

Section: Convolutional Neural Network For the Prediction Of The Biolomentioning

confidence: 99%

Meramalnet: A Deep Learning Convolutional Neural Network for Bioactivity Prediction in Structure-based Drug Discovery

Hamza

Salim

Nasser

et al. 2020

8th International Conference on Signal, Image Processing and Pattern Recognition

View full text Add to dashboard Cite

According to the principle of similar property, structurally similar compounds exhibit very similar properties and, also, similar biological activities. Many researchers have applied this principle to discovering novel drugs, which has led to the emergence of the chemical structure-based activity prediction. Using this technology, it becomes easier to predict the activities of unknown compounds (target) by comparing the unknown target compounds with a group of already known chemical compounds. Thereafter, the researcher assigns the activities of the similar and known compounds to the target compounds. Various Machine Learning (ML) techniques have been used for predicting the activity of the compounds. In this study, the researchers have introduced a novel predictive system, i.e., MaramalNet, which is a convolutional neural network that enables the prediction of molecular bioactivities using a different molecular matrix representation. MaramalNet is a deep learning system which also incorporates the substructure information with regards to the molecule for predicting its activity. The researchers have investigated this novel convolutional network for determining its accuracy during the prediction of the activities for the unknown compounds. This approach was applied to a popular dataset and the performance of this system was compared with three other classical ML algorithms. All experiments indicated that MaramalNet was able to provide an interesting prediction rate (where the highly diverse dataset showed 88.01% accuracy, while a low diversity dataset showed 99% accuracy). Also, MaramalNet was seen to be very effective for the homogeneous datasets but showed a lower performance in the case of the structurally heterogeneous datasets.

show abstract

“…Deep learning is increasingly gaining momentum in protein structure prediction (Wardah et al 2019;Senior et al 2020;Singh 2020). Deep learning strategies have also impacted on the field of drug design (Dana et al 2018). Such strategies constitutes a viable methodology which should be exploited to screen data banks of small drugs docking on cell host and viral proteases, SARS-CoV-2 surface proteins and other putative targets on the basis of structural and thermodynamic parameters, find new suitable small drug inhibitors, and explore combinations of two or more drugs acting synergistically on different targets.…”

Section: The Sars-cov-2 and Its Endogenous Proteasesmentioning

confidence: 99%

While We Wait for a Vaccine Against SARS-CoV-2, Why Not Think About Available Drugs?

Barrantes

2020

Preprint

View full text Add to dashboard Cite

At the time of reception of this article (April 2, 2020), efforts to develop a specific vaccine against SARS-Cov-2, the causative agent of the coronavirus disease 2019 (COVID-19), had just begun trial phase 1, but full validation of this and other current developments is likely to take many more months to reach completion. The ongoing pandemic constitutes a major health burden of world proportions that is also having a devastating impact on whole economies worldwide, the knock-on effects of which could be catastrophic especially in poorer countries. Alternative measures to ameliorate the impact and hamper or minimally slow down disease progression are urgently called for. This review discusses past and currently evolving data on the etiological agent of the current pandemic, SARS-CoV-2, and its host cell receptors with a view to disclosing alternative drugs for palliative or therapeutic approaches. Firstly, SARS-CoV-2 exhibits marked tropism for cells that harbor the membrane-bound metalloprotease angiotensin-converting enzyme 2 (ACE2) at their plasmalemma, predominantly in cells lining the oral cavity, upper respiratory tract, and bronchoalveolar cells, making these epithelial mucosae the most likely viral receptor cell targets and entry routes. Secondly, the crystal structures of several coronavirus spike proteins in complex with their cell host target receptors, and of SARS-Cov-2 in complex with an inhibitor, are now available at atomic resolution through X-ray diffraction and cryo-electron microscopy studies. Thirdly, viral entry of other viruses has been successfully blocked by inhibiting viral endogenous proteases or clathrin/dynamin-dependent endocytosis, the same internalization pathway followed by ACE2 and some viruses. Fourthly, the target cell-surface receptor molecules and SARS-CoV-2 possess other putative sites for drugs potentially modulating receptor activity or virus processing. A multi-pronged pharmacological approach attacking more than one flank of the viral-receptor interactions is worth considering as a front-line strategy.

show abstract

Deep Learning in Drug Discovery and Medicine; Scratching the Surface

Cited by 91 publications

References 67 publications

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

Meramalnet: A Deep Learning Convolutional Neural Network for Bioactivity Prediction in Structure-based Drug Discovery

While We Wait for a Vaccine Against SARS-CoV-2, Why Not Think About Available Drugs?

Contact Info

Product

Resources

About