One of the key lessons learned from all of the implementations that we’ve done is a need for flexibility. A customer might tell us that they need X, but it often evolves into Y over the course of the project.
Creating a system that is evolvable has its own unique challenges:
-
- users needed to be able to store data in fields that were universally recognised. It couldn’t be called X in one system and Y in another.
-
- users needed to be able to add their own field definitions.
-
- users needed to be able to create their own constrained vocabularies. Imagine a status field where the values come from a user-definable list.
Over the years we’ve seen a number of different technologies emerge to solve some of these problems. From ontologies that describe the metadata of a database, to triple-stores and graph databases to store unstructured data described by that metadata. And Query languages like SPARQL, capable of querying non-relational databases. We’ve seen attempts to expose relational databases as SPARQL endpoints using technologies db2rdf. All of these technologies have had varying degrees of success in terms of adoption.
In this post, we’ll take a look at how and why we made metadata an integral part of the foundation of our Pipeline Platform for Drug Discovery.
Metadata As The Cornerstone of Trust
The number of biotech-pharma collaborations continue to increase year over year. According to a recent report from IQVIA, “Emerging biopharma companies (EBPs) – those with an estimated expenditure on R&D of less than $200 million and less than $500 million in revenue annually – are responsible for a record 65% of the molecules in the R&D pipeline, up from less than 50% in 2016 and one-third in 2001” [1].
The Project Management Institute estimates that “half of pharmaceutical collaborations fail—not because the science falls short, but because poor alliance management leads to missteps, missed deadlines, lack of trust, and ultimately, collapse of the relationship itself” [2]
Biotech-Pharma Collaboration
Metadata can play a crucial role in building trust between biotech and pharma companies by eliminating ambiguities when sharing data. Here’s how:
-
- Common Understanding: Metadata provides a common language and understanding of the data being shared. It describes the context, structure, and meaning of the data, ensuring that both parties have the same interpretation.
-
- Data Interpretation: Metadata helps eliminate ambiguity by providing additional information about the data, such as its source, collection methods, and any transformations or modifications performed. This transparency allows for better understanding and interpretation of the data.
-
- Data Validation: Metadata enables validation of the data’s accuracy, completeness, and consistency. By providing information about data quality, metadata helps identify any potential errors or inconsistencies, building trust in the reliability of the shared data.
-
- Compliance and Regulation: Metadata assists in ensuring compliance with industry regulations and standards. By providing detailed information about data handling practices, metadata helps companies demonstrate their adherence to regulatory requirements, fostering trust and confidence.
-
- Data Lineage and Provenance: Metadata provides a record of the data’s lineage, including its origin, transformations, and any intermediate steps. This transparency helps trace the data back to its source, ensuring accountability and traceability, which are essential for building trust.
-
- Facilitating Collaboration: Metadata facilitates seamless collaboration between biotech and pharma companies by providing a shared understanding of the data. It enables effective communication, reduces misunderstandings, and allows for efficient data integration and analysis.
Biotech CRO Collaboration
In today’s resource-constrained environment, biotech companies rely on Clinical Research Organisations (CROs) more than ever to perform experiments at a scale that is more capital efficient and scalable.
“…the outsourcing trend for drug discovery as a whole is predicted to continue to grow over the next decade by 5−10% per year, and the drug discovery chemistry market is predicted to grow to $17.7 Bn by 2025.”
Source: Drug Discovery: Collaborations between Contract Research Organizations and the Pharmaceutical Industry – ACS Med Chem Letters
However, those relationships also come with their own risks. Similar to the risks found in biotech-pharma collaborations, biotech companies and their CRO partners must be on the same page when it comes to exchanging data. They need a common understanding of what data will be provided and how it was generated.
Internal Communication
As much as metadata affects external communication, it’s just as important for clear internal communication as well. Drug discovery is a team sport, and most of the players have different academic backgrounds, where certain terms will have very domain specific meanings. The Irish playwright, George Bernard Shaw once remarked that Brits and Americans were a people divided by a common language, and the same can be said about biologists and chemists.
Metadata Everywhere All At Once
In Pipeline, metadata is ubiquitous. At the core of the application is the Field Library which contains the definitions of fields imported from industry standard ontologies such as the Allotrope ontology, the EDAM ontology, NCI Metathesaurus and BioAssay Ontology (BAO). In fact, any ontology found in the NCBO’s BioPortal can be imported and used in Pipeline.
Field Library
When importing assay data, the fields are mapped to these vocabularies to disambiguate them. Since the fields in the Field Library definitions often contain validation rules, the system uses those rules to validate the data being imported into the system.
When exporting data, the data and metadata are included in the export file. When you hover over a column header, the definition of the field appears as a comment in the file. This same metadata is also exported, every time a template file is generated, thus ensuring that the template file itself describes the data that it contains in terms that are unambiguous to the person collecting the assay data.
In calculations, the system maps the fields in the field library to the input and output parameters of the calculation. Thereby eliminating another potential source of error.
Lastly, the user interface itself contains metadata. If the user hovers over a column in a table, the metadata for that column appears as a blue popup (shown below). If the user is entering data for a given row of that table, the help text for the field will automatically appear. And any validation criteria found in the Field Library will be applied to the field.
Not Just The What But The How
But it’s not really sufficient to simply describe the data that you’re sharing with another person. It’s important for them to understand HOW the data were collected, and processed. In other words, it’s not just the WHAT that’s important but the HOW. To facilitate this, the Pipeline system has a built-in Protocol Management module that provides detailed information about the procedure and materials used to collect the data.
If a calculation was used, then the parameters are mapped to entries in the Field Library. Moreover, if the calculation was provided as a script, then the script itself is accessible and visible, making it easier to understand how the data were processed and tested.
[1] Global Trends in R&D 2022 – IQVIA
—-
If you’d like to find out more about Pipeline, contact us at info@aspen.bio for a demonstration