Maximizing the Value of Structural Data

As we explained in our previous article, the number of biomolecular structures available for computational analysis is increasing daily, which, for SBDD translates to ever-greater opportunities for discovery.

But wait, is that actually true?

Sort of... The number of structures alone does not dictate their value. The value of a structure resides in the information it contains and how useful that information is.

We could argue that a structure is useful if it brings new insights. If we are studying protein folds, probably with a few structures we could have enough, but if we are in the context of drug discovery, the more ligands crystallized the more information we gain (of course if the ligands are different between each other). Therefore, not every new structure will bring value.

What really will determine the value of a structure and its usefulness is the scope:

A structure can be useful to answer a very specific question in a particular project: e.g. did the new methyl group alter the binding conformation of the ligand? The same structure under a different scope could be useful to study binding site flexibility, to build pharmacophore models or to study protein dynamics amongst many other examples. In each case, the information we extract and consider will differ and thus the value of the structure will depend on who is studying it and for what purpose.

This means that a structure that no longer brings you any value, can have a huge value for your colleague, now or in the future. It will certainly have value in any statistical study. Therefore, the value can change over time and in different contexts. The common invariable point is the information within the structure.

For that reason, to ensure the usefulness of all the structures produced, we must keep track of the information they contain in a well-organized, easily accessible manner and through time.

Moreover, facilitating the collaboration between colleagues with different backgrounds will booster, even more, the value of each structure.

Principles for Optimal Data Exploitation

At Discngine, we have identified four principles that we consider cardinal for maximizing the value of computational protein structures for Drug Discovery projects.

Store relevant information, not files.

Because it is the information that matters and not the number, we cannot simply put the new structure files in a folder. The information that is relevant in our scope of Drug Discovery, needs to be extracted, organized and well presented, to expose the value.

For instance, in order to be able to assess side effects, one might be interested in searching for similar binding sites to your target protein. To do so, we need information about each binding site in the existing structures.

Data must be easily accessible and searchable for all users,

meaning that interfaces, terms, and functions should be intuitive and that all structures should accessible in the same place.

For instance, software users should be able to quickly navigate through full protein structures or specific regions, such as binding pockets, across large sets of data (i.e. hundreds of proteins or more). Chemists should easily find docking models produced by their modeler colleagues.

Structure-Based Idea Generation should be popularized.

Bench biologists and chemists should have the possibility to take advantage of protein-ligand binding 3D spatial data.

For example, researchers without expertise in Computational Biology or Chemistry should be able to effortlessly scan large sets of protein structures and pockets, even outside their desired target class, to gain inspiration, especially in unexpected places.

Structural data needs to integrate seamlessly with other types of information and software,

such as proprietary reports, screening results and personal annotations contributed by team members from diverse backgrounds, including Protein Science, Structural Biology, Computational Chemistry, and Medicinal Chemistry. The ideal applications should integrate well with the existing ecosystem of computational modeling tools and collaborative drug design platforms.

However, this can only be possible if all the structural data are centralized in one location that makes data navigation and collaborative analysis easy.

What is 3decision®?

3decision® is our attempt at addressing the critical points we identified in order to maximize the use and usefulness of the vast structural data we have at hand. It is a protein structure repository with a collaborative web-based interface that allows your whole Drug Discovery team to centralize, visualize, analyze and annotate protein and ligand structural data for collaborative ideation and decision-making.

 Each new protein structure, either experimental or in-silico, from public or internal sources, can be uploaded. They are processed to extract as much information as we deem appropriate in the context of SBDD to render it useful. Protein sequences, annotations, binding sites, ligands, interactions - all are characterized, stored and coupled with smart and interactive visualizations. This allows you to quickly compare or analyze sets of structures within your project or explore and browse the vast amount of information using advanced algorithms, like pocket, sequence or ligands similarities.

 The aim is to facilitate and speed up the data exploration and hypothesis generation & validation in a multidisciplinary working environment.

 The integration in an already existing ecosystem of computational tools is possible thanks to 3decision’s API that allows data to register and retrieve seamlessly. Expert users that work with sophisticated modeling tools and algorithms need to be able to pull and push data easily.

3decision in practice

 Some small examples are worth more than a thousand words so, how does 3decision® help you in a drug design project?

 Case 1: Early stages of a new project

• Search and download existing structures on the target and homologous proteins.

• Find druggable binding sites.

• Identify potential off-targets.

• Find existing binding chemical matter for your target or related proteins (homologous, off-targets).

• Download superimposed structures to use them in modeling tools to build pharmacophores and/or run virtual screening campaigns.

• No structures for the target yet? Search for homologous proteins and identify the best templates to build homology models from with some visual aid.

• Search similar proteins and extract crystallization parameters.

• Modelers might run molecular dynamics with their favorite software and upload different model structures that represent the conformational flexibility of the protein.

• Developing in-vitro, in-vivo assays? Check mutations in the binding site between species in 3D to avoid undesired surprises.

Case 2: Lead optimization loop

• Upload a new structure: new holo-structure resolved, new docking model, new homology model.

•  Prepare a 3D scene comparing this new structure with the previous ones in the project to highlight differences with potential interest. E.g: new subpocket opens giving room to add new functional groups to the ligand.

• Compare against off-target structures to reveal mutations in the binding site that could be exploited to gain specificity.

• Share the scene with your colleagues with a simple URL link that will serve as a basis for discussion to make decisions.

• Use the integrated 2-clicks docking feature to quickly test the fitting of the new ideas in the binding site. If more detailed and precise dockings need to be produced, the modeler can take care of this step and upload the models into 3decision restarting the loop.

• Some new molecules will be synthesized at some point and new crystal structures will be resolved, restarting this way the loop. 

Market recognition

Prestigious Biotech and Pharma companies such as Abbvie and Lundbeck are already using 3decision® to accelerate Structure-Based Drug Design.

 In fact, 3decision® was originally co-developed by Discngine through a partnership with Abbvie, who received the Innovative Practices Award at BioIT World 2019 for their use of the software.

If 3decision® sounds like the right platform for your Drug Discovery team to tackle the challenges and demands of the growing mountain of public and proprietary structural data, then contact Discngine today to learn more and sign up for your free trial.

3decision’s blog has been moved!