Spatiotemporal identification of druggable binding sites using deep learning

Kozlovskii, Igor; Popov, Petr

doi:10.1101/2020.02.20.952309

Cited by 5 publications

(8 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To detect binding site satisfying those rationales, we applied BiteNet [22], a deep learning approach for spatiotemporal identification of druggable binding sites, to the 10µs MD simulation trajectories by D.E. Shaw Research for the Spike structure in the closed and prefusion states [17] (see V A).…”

Section: Resultsmentioning

confidence: 99%

“…Typical obstacles in binding site identification include pitfalls related to the i) flexibility, ii) druggability, iii) accessibility and iv) mutability of a protein. Firstly, protein flexibility is crucial in drug discovery [42], and a binding site may be present or absent in a given three-dimensional structure; hence, there is a risk of overlooking a relevant binding site or detecting a fleeting irrelevant binding site [22]. Secondly, not every detected binding site is 'druggable', meaning that one can make a drug that modulates protein function upon binding [3].…”

Section: Introductionmentioning

confidence: 99%

“…In this study, we identified a vulnerable region in the Spike trimer structure that could be used to allosterically inhibit RBD-PD interactions by preventing closed to open conformational transition of Spike. We analyzed the long-range molecular dynamics (MD) trajectories of Spike using a spatiotemporal deep learning-based approach BiteNet [22] and selected conformation-and oligomer-specific cryptic binding site based on the putative mechanism of action and structure-based criteria. Namely, the detected binding site is formed by two spike subunits, it shares amino acid residues with the RBD of one subunit, and it is present in the closed, but not the open conformation of Spike.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Rationalizing Viral Drug Target Identification Using Computational Approaches : The SARS-CoV-2 Spike Glycoprotein S Case Study

Popov

Buslaev²,

Kozlovskii³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

<div><div><div><p>COVID-19 emphasized the need for fast reaction tools to fight global biological threats such as viruses. Rapid drug discovery is one of the strategies for efficient social response. The success of a drug discovery campaign critically depends on the selected drug target, and the wrong target nullifies all the efforts to develop a drug. Viral drug target identification is a challenging problem, and computational methods can reduce the number of candidate targets. Here we present a structure-based approach to identify vulnerable regions in viral proteins that comprise drug binding sites. To detect promising binding sites, we take into account protein dynamics, accessibility, and mutability of the binding site, coupled with the putative mechanism of action of a drug. Applying to the SARS-CoV-2 Spike Glycoprotein S, we observed conformation- and oligomer-specific glycan-free binding site that is proximal to the receptor binding domain and comprises topologically important amino acid residues. Molecular dynamics simulations of Spike in complex with drug-like molecules docked into the binding sites revealed shifted equilibrium towards the inactive conformation compared to the ligand-free simulations. Small molecules targeting this binding site could prevent the closed-to-open conformational transition of the Spike protein, thus, allosterically inhibit the interaction with the human angiotensin-converting enzyme 2 receptor.</p></div></div></div>

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Rationalizing Viral Drug Target Identification Using Computational Approaches : The SARS-CoV-2 Spike Glycoprotein S Case Study

Popov

Buslaev²,

Kozlovskii³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The DL prediction has been focused on this region of Bcl-2, as the goal of this approach is to detect holo-like conformations i.e favorable conformations for ligand binding and not to detect the binding region on the whole protein surface. Figure 9 [36,37]. Likewise, the interactability was monitored along the same simulation and the RMSD of the binding site residues was compared to a holo HD structure of Bcl-2 bound to Bax (pdb 2xa0) Figure 9 (Right panel).…”

Section: Case Study: Bcl-2 As Therapeutic Targetmentioning

confidence: 99%

InDeep : 3D fully convolutional neural networks to assist in silico drug design on protein-protein interactions

Mallet

Ruano

Moine-Franel

et al. 2021

Preprint

View full text Add to dashboard Cite

MotivationProtein-protein interactions (PPIs) are key elements in numerous biological pathways and the subject of a growing number of drug discovery projects including against infectious diseases. Designing drugs on PPI targets remains a difficult task and requires extensive efforts to qualify a given interaction as an eligible target. To this end, besides the evident need to determine the role of PPIs in disease-associated pathways and their experimental characterization as therapeutics targets, prediction of their capacity to be bound by other protein partners or modulated by future drugs is of primary importance.ResultsWe present InDeep, a tool for predicting functional binding sites within proteins that could either host protein epitopes or future drugs. Leveraging deep learning on a curated data set of PPIs, this tool can proceed to enhanced functional binding site predictions either on experimental structures or along molecular dynamics trajectories. The benchmark of InDeep demonstrates that our tool outperforms state of the art ligandable binding sites predictors when assessing PPI targets but also conventional targets. This offers new opportunities to assist drug design projects on PPIs by identifying pertinent binding pockets at or in the vicinity of PPI interfaces.AvailabilityThe tool is available on GitHub3 along with a PyMol plugin for visualization. Predictions of InDeep can be consulted at iPPI-DB4

show abstract

“…In consideration of these limitations, some researchers have started to conduct interaction studies from the perspective of structure feature mining. Kozlovskii et al used the deep learning method to dig the structure of known protein-ligand complexes and achieved satisfactory results in certain druggable binding site identification (Kozlovskii and Popov 2020). It can be seen that all these works without exception encounter limitations so that they can only focused on a certain mode of interactions like docking and binding or certain interaction materials like metal ions, small molecules and ligands.…”

Section: Introductionmentioning

confidence: 99%

TMPs Interaction Domain: A Discovery of Structural Universality Among TMPs Interaction Sites Using Deep Learning Method

Wang

Ming

et al. 2021

Preprint

View full text Add to dashboard Cite

Transmembrane proteins (TMPs) serve as important drug targets and accounts for nearly half of the drugs currently available in the market. Research into TMPs interactions and their structural basis will provide key information for drug research and new drug development. Based on previous works like a binding pocket or binding site, our main purpose in this study is to find whether the structural universality (Interaction Domain) exists in all kinds of TMPs interaction regions through a computational approach. After implementing the experiments using our 3D deep learning model and achieve the Matthews correlation coefficient (MCC) of 0.36, we found strong evidence for the existence of the structural basis among TMPs interaction regions. That means those regions, or we call them interaction domains, are structural specific distinguishing to the domains without any interaction. According to this, this work provides a new theoretical basis for TMPs interaction research and can greatly boost the development of the drug industry.

show abstract

Spatiotemporal identification of druggable binding sites using deep learning

Cited by 5 publications

References 51 publications

Rationalizing Viral Drug Target Identification Using Computational Approaches : The SARS-CoV-2 Spike Glycoprotein S Case Study

Rationalizing Viral Drug Target Identification Using Computational Approaches : The SARS-CoV-2 Spike Glycoprotein S Case Study

InDeep : 3D fully convolutional neural networks to assist in silico drug design on protein-protein interactions

TMPs Interaction Domain: A Discovery of Structural Universality Among TMPs Interaction Sites Using Deep Learning Method

Contact Info

Product

Resources

About