Abstract:Artificial intelligence can train the related known drug data into deep learning models for drug design, while classical algorithms can design drugs through established and predefined procedures. Both deep learning and classical algorithms have their merits for drug design. Here, the webserver WADDAICA is built to employ the advantage of deep learning model and classical algorithms for drug design. The WADDAICA mainly contains two modules. In the first module, WADDAICA provides deep learning models for scaffol… Show more
“…The performance of the MLR model is expressed in terms of R 2 , which was found to be 0.799 signifying that ∼ 80% of the data fit the regression model. Cpd 9 with the highest binding affinity value from the WADDAICA server [36] , was also predicted to have the highest binding affinity by the MLR model. Indeed, the MLR model agreed with the binding affinity predictions of the WADDAICA webserver but it should be noted that many of the variables used in the MLR model were obtained from other prediction algorithms – such as binding affinity, logP and TPSA.…”
Section: Resultsmentioning
confidence: 94%
“…The full SwissADME parameters are provided in the supplementary material for Cpds 1–9 Fig. 6 a) Protein-ligand interaction fingerprint map of the 9 docked compounds with GPR120S; Green – (HB) Hydrogen bond; Yellow – (HP) Hydrophobic interactions; Grey – No interactions; b) Heatmap of protein–ligand binding affinities of snapshots extracted from the 100 ns MD simulation trajectory and scored by webserver WADDAICA [36] ; c) Protein-ligand interaction fingerprint map [46] plotting the number of interactions of the residue with the ligands, shows compounds 1, 7 and 9 with conserved W277 and N313 H-bond interactions over the period of 100 ns MD production runs. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)…”
Section: Resultsmentioning
confidence: 99%
“… Fig. 8 Scatter plot of a multiple linear regression model of the contribution of physicochemical descriptors –molecular weight, number of H-bond acceptors and H-bond donors, TPSA and logP from SwissADME [37] over the WADDAICA [36] binding affinity for the 9 compounds. The full SwissADME parameters are provided in the supplementary material for Cpds 1–9.…”
Section: Resultsmentioning
confidence: 99%
“…The overall stability of the simulated systems was also checked with respect to temperature, pressure, and potential energy of the systems to check thermodynamic equilibrium during the production simulation runs, confirming the convergence of individual trajectories. The WADDAICA webserver was used to compute the protein–ligand binding free energies of protein–ligand snapshots extracted from the MD trajectory every 5 ns using the Binding Affinity by AI module of WADDAICA [36] . The protein–ligand interaction plots from the MD trajectory were generated using Molecular-dynamics-Interaction-plot (available at https://github.com/tavolivos/Molecular-dynamics-Interaction-plot ).…”
Section: Methodsmentioning
confidence: 99%
“…The protein–ligand interaction plots from the MD trajectory were generated using Molecular-dynamics-Interaction-plot (available at https://github.com/tavolivos/Molecular-dynamics-Interaction-plot ). Multiple linear regression (MLR) was used to correlate the physicochemical descriptors (obtained from SwissADME [37] ) with the binding affinity predictions from WADDAICA [36] . The ChemMine tools webserver was used to develop a structural similarity clustering scatterplot [38] .…”
“…The performance of the MLR model is expressed in terms of R 2 , which was found to be 0.799 signifying that ∼ 80% of the data fit the regression model. Cpd 9 with the highest binding affinity value from the WADDAICA server [36] , was also predicted to have the highest binding affinity by the MLR model. Indeed, the MLR model agreed with the binding affinity predictions of the WADDAICA webserver but it should be noted that many of the variables used in the MLR model were obtained from other prediction algorithms – such as binding affinity, logP and TPSA.…”
Section: Resultsmentioning
confidence: 94%
“…The full SwissADME parameters are provided in the supplementary material for Cpds 1–9 Fig. 6 a) Protein-ligand interaction fingerprint map of the 9 docked compounds with GPR120S; Green – (HB) Hydrogen bond; Yellow – (HP) Hydrophobic interactions; Grey – No interactions; b) Heatmap of protein–ligand binding affinities of snapshots extracted from the 100 ns MD simulation trajectory and scored by webserver WADDAICA [36] ; c) Protein-ligand interaction fingerprint map [46] plotting the number of interactions of the residue with the ligands, shows compounds 1, 7 and 9 with conserved W277 and N313 H-bond interactions over the period of 100 ns MD production runs. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)…”
Section: Resultsmentioning
confidence: 99%
“… Fig. 8 Scatter plot of a multiple linear regression model of the contribution of physicochemical descriptors –molecular weight, number of H-bond acceptors and H-bond donors, TPSA and logP from SwissADME [37] over the WADDAICA [36] binding affinity for the 9 compounds. The full SwissADME parameters are provided in the supplementary material for Cpds 1–9.…”
Section: Resultsmentioning
confidence: 99%
“…The overall stability of the simulated systems was also checked with respect to temperature, pressure, and potential energy of the systems to check thermodynamic equilibrium during the production simulation runs, confirming the convergence of individual trajectories. The WADDAICA webserver was used to compute the protein–ligand binding free energies of protein–ligand snapshots extracted from the MD trajectory every 5 ns using the Binding Affinity by AI module of WADDAICA [36] . The protein–ligand interaction plots from the MD trajectory were generated using Molecular-dynamics-Interaction-plot (available at https://github.com/tavolivos/Molecular-dynamics-Interaction-plot ).…”
Section: Methodsmentioning
confidence: 99%
“…The protein–ligand interaction plots from the MD trajectory were generated using Molecular-dynamics-Interaction-plot (available at https://github.com/tavolivos/Molecular-dynamics-Interaction-plot ). Multiple linear regression (MLR) was used to correlate the physicochemical descriptors (obtained from SwissADME [37] ) with the binding affinity predictions from WADDAICA [36] . The ChemMine tools webserver was used to develop a structural similarity clustering scatterplot [38] .…”
The ability of alcohol extracted constituents from the leaves of Clerodendrum paniculatum to inhibit SARS-CoV-2 was investigated through various computational methods like docking, molecular dynamic simulations and pharmacokinetic predictions. Of the various active constituents, quercetin was identified as a potent inhibitor that can bind strongly to the active site of main protease (M pro ) and spike protein of the virus with respective binding free energy of À 41.07 and À 40.76 kcal mol À 1 . Molecular dynamic simulations also supported the binding interactions by the presence of strong hydrogen bonding interactions with the key residues in the binding pocket of the target protein. The results were ascertained experimentally by evaluating the inhibition potential of the extract against spike and M pro proteins of SARS-CoV-2.
De novo drug design is a stationary way to build novel ligands in the confined pocket of receptor by assembling the atoms or fragments, while molecular dynamics (MD) simulation is a dynamical way to study the interaction mechanism between the ligands and receptors based on the molecular force field. De novo drug design and MD simulation are effective tools for novel drug discovery. With the development of technology, deep learning methods, and interpretable machine learning (IML) have emerged in the research area of drug design. Deep learning methods and IML can be used further to improve the efficiency and accuracy of de novo drug design and MD simulations. The application summary of deep learning methods for de novo drug design, MD simulations, and IML can further promote the technical development of drug discovery. In this article, two major workflow methods and the related components of classical algorithm and deep learning are described for de novo drug design from a new perspective. The application progress of deep learning is also summarized for MD simulations. Furthermore, IML is introduced for the deep learning model interpretability of de novo drug design and MD simulations. Our paper deals with an interesting topic about deep learning applications of de novo drug design and MD simulations for the scientific community.
This article is categorized under:
Data Science > Chemoinformatics
Data Science > Artificial Intelligence/Machine Learning
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
