The capitalized cost of discovering a new drug can be approximately estimated at more than a billion dollars and often a decade of research starting from the identification of a new target up to completing clinical studies. This calls for powerful means for accurately studying and designing new drugs. This issue has been traditionally tackled by experimental procedures to estimate the protein-ligand binding thermodynamics, kinetics and in general in vivo properties of a drug compound. In the last 50 years, however, computational means have increasingly gained a prominent role in the design process of drugs as compounds and proteins can be either accurately simulated, for instance, by physically-driven approaches (e.g. molecular dynamics), or machine learning based black box-type tools, which both can be employed to predict physical observables.
In this collection of articles, we would like to gather contributions in the field of molecular dynamics and machine learning aimed at dissecting the drug discovery issue. Molecular dynamics represents the now classical, physics-based approach and it has been found to be capable, especially via enhanced sampling methods, to estimate thermodynamics and kinetics of protein-ligand binding. Additionally, molecular dynamics approaches can generate putative binding and unbinding trajectories and thus providing means of mechanistic interpretations at the atomistic level. From another side, machine learning has first historically contributed with Quantitative Structure Activity Relationships (QSAR) approaches and now is emerging via deep learning and generative techniques as a further possibility, not only to predict drug properties, but also, combined with molecular dynamics, to devise efficient biasing strategies, analysis methods and data driven collective variables. In this scenario, this collection would like to bring together contributions of both black and white box modeling also encouraging works at the intersection that leverage the features of the two approaches.
Topics may include but are not limited to:
• Molecular dynamics simulations of protein ligand binding/unbinding
• Enhanced sampling simulations for drug discovery
• Prediction thermodynamics and kinetics of protein-ligand binding via molecular dynamics or machine learning
• Machine learning analysis of molecular dynamics simulations for drug discovery, particularly deep learning methods
• New methodologies for enhanced sampling or machine learning for rare events (e.g. ligand unbinding)
• Role of water molecules in drug binding
• Drugs metabolism
• New therapeutic (e.g. PROTAC) simulations and modeling
-------------------------------------------------------------------------------
Dr. Sergio Decherchi and Dr. Andrea Cavalli are co-founders of BiKi Technologies s.r.l. - a company that commercializes a Molecular Dynamics-based software suite for drug discovery. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
The capitalized cost of discovering a new drug can be approximately estimated at more than a billion dollars and often a decade of research starting from the identification of a new target up to completing clinical studies. This calls for powerful means for accurately studying and designing new drugs. This issue has been traditionally tackled by experimental procedures to estimate the protein-ligand binding thermodynamics, kinetics and in general in vivo properties of a drug compound. In the last 50 years, however, computational means have increasingly gained a prominent role in the design process of drugs as compounds and proteins can be either accurately simulated, for instance, by physically-driven approaches (e.g. molecular dynamics), or machine learning based black box-type tools, which both can be employed to predict physical observables.
In this collection of articles, we would like to gather contributions in the field of molecular dynamics and machine learning aimed at dissecting the drug discovery issue. Molecular dynamics represents the now classical, physics-based approach and it has been found to be capable, especially via enhanced sampling methods, to estimate thermodynamics and kinetics of protein-ligand binding. Additionally, molecular dynamics approaches can generate putative binding and unbinding trajectories and thus providing means of mechanistic interpretations at the atomistic level. From another side, machine learning has first historically contributed with Quantitative Structure Activity Relationships (QSAR) approaches and now is emerging via deep learning and generative techniques as a further possibility, not only to predict drug properties, but also, combined with molecular dynamics, to devise efficient biasing strategies, analysis methods and data driven collective variables. In this scenario, this collection would like to bring together contributions of both black and white box modeling also encouraging works at the intersection that leverage the features of the two approaches.
Topics may include but are not limited to:
• Molecular dynamics simulations of protein ligand binding/unbinding
• Enhanced sampling simulations for drug discovery
• Prediction thermodynamics and kinetics of protein-ligand binding via molecular dynamics or machine learning
• Machine learning analysis of molecular dynamics simulations for drug discovery, particularly deep learning methods
• New methodologies for enhanced sampling or machine learning for rare events (e.g. ligand unbinding)
• Role of water molecules in drug binding
• Drugs metabolism
• New therapeutic (e.g. PROTAC) simulations and modeling
-------------------------------------------------------------------------------
Dr. Sergio Decherchi and Dr. Andrea Cavalli are co-founders of BiKi Technologies s.r.l. - a company that commercializes a Molecular Dynamics-based software suite for drug discovery. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
