Biologic drugs including mainly monoclonal antibodies and also fusion proteins are presently the mainstream trend in biopharmaceutical research for treating major illnesses such as cancer and infectious diseases. Benefitting from sophisticated methods for detailed structural characterization at the molecular level, in silico molecular modeling structure-based approaches have emerged as a key toolbox employed in the biologic drug development pipeline.
Up until now, classical molecular modeling methods based on molecular mechanics force-fields and empirical linear models have matured to a level of routine implementation during hit optimization flows in most lead development campaigns. There is no doubt, however, that the accuracy of these in silico predictions still lacks in certain key areas, for example binding affinity or aggregation predictions to name a few, where additional refinements will be much welcomed.
At the same time, with the fast-paced infusion of artificial intelligence in many research areas impacting everyday life, we are witnessing a new chapter being written in drug design. Several nonlinear models ranging from machine learning to unsupervised learning with deep neural networks, which feed on still relatively small but growing biological and structural datasets, are emerging as formidable competing alternatives to classical methods. There are also growing attempts to pair structure-based computational methods with wet lab library screening and display technology for more efficient antibody hit-to-lead optimization. While most of the structure-based applications have tackled the relatively easier problem of optimising existing antibody candidates, there are signs that de novo discovery of antibody hits entirely by structure-based computational methods will be feasible and may become a norm in the not-so-distant future.
This collection thus seizes a high time, when the accuracy limits can be pushed for classical methods, and solid foundations can be established for artificial intelligence methods. Together, towards a common goal, these methods can robustly exploit protein structure information to benefit the entire biologics drug development pipeline, from the molecular design, through property optimization, and to scaled-up manufacturability. Hence, this Research Topic welcomes the submission of Original Research, Reviews or Perspective articles on all aspects related to the use of structure-based molecular modeling methods that support all stages along the biologic drug design, discovery, optimization and development pipeline. This includes but is not limited to:
· Prediction of binding affinity (e.g., antibody-antigen affinity maturation)
· Molecular engineering of specificity (e.g., multiple targets, tissue specificity)
· In silico de-risking of developability attributes (e.g., immunogenicity, aggregation, stability)
· Structure-based modulation of immune effector functions and pharmacokinetics
· De novo in silico design of antibodies
The breadth of structure-based computational methods encompassed by this Research Topic is fairly broad and includes but is not limited to:
· Advances with classical molecular modeling methods (linear, empirical, force-field)
· Recent forays into AI methods for drug design (non-linear, machine learning, deep neural networks)
· 3D structure-based methods
· Protein-protein docking
· Sequence-based methods
· Structure-function database assembly, curation and mining
In silico studies submitted to this Research Topic are encouraged to make use of experimental data for training computational models and/or validation of predictions, either generated in the respective studies or taken from the relevant published literature.
---------
Dr. Sandeep Kumar is an employee of Boehringer Ingelheim and holds patents related to this Research Topic
Biologic drugs including mainly monoclonal antibodies and also fusion proteins are presently the mainstream trend in biopharmaceutical research for treating major illnesses such as cancer and infectious diseases. Benefitting from sophisticated methods for detailed structural characterization at the molecular level, in silico molecular modeling structure-based approaches have emerged as a key toolbox employed in the biologic drug development pipeline.
Up until now, classical molecular modeling methods based on molecular mechanics force-fields and empirical linear models have matured to a level of routine implementation during hit optimization flows in most lead development campaigns. There is no doubt, however, that the accuracy of these in silico predictions still lacks in certain key areas, for example binding affinity or aggregation predictions to name a few, where additional refinements will be much welcomed.
At the same time, with the fast-paced infusion of artificial intelligence in many research areas impacting everyday life, we are witnessing a new chapter being written in drug design. Several nonlinear models ranging from machine learning to unsupervised learning with deep neural networks, which feed on still relatively small but growing biological and structural datasets, are emerging as formidable competing alternatives to classical methods. There are also growing attempts to pair structure-based computational methods with wet lab library screening and display technology for more efficient antibody hit-to-lead optimization. While most of the structure-based applications have tackled the relatively easier problem of optimising existing antibody candidates, there are signs that de novo discovery of antibody hits entirely by structure-based computational methods will be feasible and may become a norm in the not-so-distant future.
This collection thus seizes a high time, when the accuracy limits can be pushed for classical methods, and solid foundations can be established for artificial intelligence methods. Together, towards a common goal, these methods can robustly exploit protein structure information to benefit the entire biologics drug development pipeline, from the molecular design, through property optimization, and to scaled-up manufacturability. Hence, this Research Topic welcomes the submission of Original Research, Reviews or Perspective articles on all aspects related to the use of structure-based molecular modeling methods that support all stages along the biologic drug design, discovery, optimization and development pipeline. This includes but is not limited to:
· Prediction of binding affinity (e.g., antibody-antigen affinity maturation)
· Molecular engineering of specificity (e.g., multiple targets, tissue specificity)
· In silico de-risking of developability attributes (e.g., immunogenicity, aggregation, stability)
· Structure-based modulation of immune effector functions and pharmacokinetics
· De novo in silico design of antibodies
The breadth of structure-based computational methods encompassed by this Research Topic is fairly broad and includes but is not limited to:
· Advances with classical molecular modeling methods (linear, empirical, force-field)
· Recent forays into AI methods for drug design (non-linear, machine learning, deep neural networks)
· 3D structure-based methods
· Protein-protein docking
· Sequence-based methods
· Structure-function database assembly, curation and mining
In silico studies submitted to this Research Topic are encouraged to make use of experimental data for training computational models and/or validation of predictions, either generated in the respective studies or taken from the relevant published literature.
---------
Dr. Sandeep Kumar is an employee of Boehringer Ingelheim and holds patents related to this Research Topic