|Home | About | Journals | Submit | Contact Us | Français|
When we look at the rapid growth of scientific databases on the Internet in the past decade, we tend to take the accessibility and provenance of the data for granted. As we see a future of increased database integration, the licensing of the data may be a hurdle that hampers progress and usability. We have formulated four rules for licensing data for open drug discovery, which we propose as a starting point for consideration by databases and for their ultimate adoption. This work could also be extended to the computational models derived from such data. We suggest that scientists in the future will need to consider data licensing before they embark upon re-using such content in databases they construct themselves.
Public online databases  supporting life sciences research have become valuable resources for researchers depending on data for use in cheminformatics, bioinformatics, systems biology, translational medicine, and drug repositioning efforts, to name just a few of the potential end user groups. Worldwide funding agencies (governments and not-for-profits) have invested in public domain chemistry platforms. In the United States these include PubChem , ChemIDPlus , and the Environmental Protection Agency's ACToR , while the United Kingdom has funded ChEMBL  and ChemSpider , among others, and new databases continue to appear annually .
We have argued recently that the data quality contained within many of these databases is suspect  and scientists should consider issues of data quality  when using these resources. By assimilating various data sources together and meshing data on drugs, proteins, and diseases, these various databases and network and computational methods may be useful to accelerate drug discovery efforts. The development of related cheminformatics platforms or derived models without care given to data quality is a poor strategy for long-term science  as errors become perpetuated in additional databases. There is real evidence that the integration of large, heterogeneous sets of databases and other types of content is “unreasonably effective” at accelerating the conversion of data into knowledge . This implies the need for technical and semantic work to bring databases together that were never designed for interoperability , which is in itself a significant task , .
The availability of data for download and reuse is an important offering to the community, as these data may be used for the purpose of modeling to develop prediction tools . In addition, data can be ingested into internal systems inside pharmaceutical companies to mesh with their existing private data , including in the expanding Linked Open Data cloud or in freely available online databases, and can be downloaded and used to enhance their content and to establish linking between data. The Open PHACTS project ,  utilizes a semantic web approach to integrate chemistry and biology data across a myriad of data sources, including for chemistry ChEBI, ChEMBL, and DrugBank, and for biology UniProt, Wikipathways, and many others. The chemical structure representations are obtained from ChemSpider, which has previously imported the chemical databases and standardized according to their data model and are making the data available as open data to the project. Many of the primary online databases already have multiple links to external systems. This linking may be achieved by using available database services to form transitory links in by, for example, using a chemical representation such as an InChI  to probe an application programming interface, search for the compound, and generate the linking URL in real time. Commonly, however, the links are more permanent in nature and are generated by downloading data from the various data sources, depositing a subset of the data (generally the chemical compound and associated database identifier), and using the particular database URL structure to form permanent links. This act of download and deposition of multiple data sources is commonly mixing the various licenses, if licenses are even declared, which, in many cases, they are not.
In some ways, there are analogous difficulties in the exchange of computational models like quantitative structure activity relationship (QSAR) datasets —while there are efforts to standardize how the data and models are stored, queried, and exchanged, there has been little consideration of licenses required to enable making the sharing of open source models a reality . Similarly, one could consider the creation of maps of disease and how they are shared and reused  in the same manner.
The potential legal fragility of knowledge products derived from online databases with poorly understood licensing for each of the databases is a real problem, and one that will only increase in severity over time. This realization is not novel; indeed, the chemical blogosphere has been host to many discussions regarding the need for clear data licensing definitions on chemistry-related data. Many scientists likely echo these comments, but we will provide some examples. In particular, Peter Murray-Rust  espouses the value of “open data”  to the scientific discovery process and encourages clear licensing of all chemistry data according to Open Knowledge Definition (OKD)  and the Panton Principles .
Herein we provide an extensive background to the intellectual property around data and databases in the sciences involved in drug discovery, those of biology, chemistry, and related fields, as well as discussion of open data licensing, openness, and open license limitations (Text S1). More importantly, we provide a set of rules that practitioners might apply when making data or databases available via the Internet or mobile apps . Our ultimate goal is to illuminate the legal fragility of the database ecosystem in the drug discovery sciences, and to initiate a conversation about creating best practices.
We suggest based on our analysis of the current data situation (Text S1) the ideal is to use strong default rules for openness. From a copyright and database rights perspective, the public domain gives the most clarity and should be the default setting for data deposit, although it may not always be achievable. Understanding this is vital, because it sets the bar at the right height. Justifications for additional controls should be subject to argument—one often finds those controls are unnecessary when the discussion is framed this way.
It is also important to avoid noncommercial or share-alike approaches whenever possible. These are attractive terms to many data providers, but create significant barriers to interoperability. Noncommercial data might be incompatible for researchers at a pharmaceutical company, even to run a simple web-based query. It is important to realize data under a share-alike license from one entity is probably not combinable with data under a share-alike license from another entity (this lack of interoperability kept Creative Commons licensed images out of Wikipedia for years, and is not one we wish to introduce into the ecosystem again!).
Thus, we propose the following simple rules for developing data licensing approaches inside scientific projects.
As a general rule, these four simple rules should allow us to build a more stable data and model sharing ecosystem while we live with some uncertainties until the courts rule on where the line of property stops and starts. We can't wait for the certainty to emerge, but we also want our systems to work when the courts do finally rule on issues such as where data and metadata stop and start, where copyright attaches, how data rights really affect re-use, and what it means to move towards a “cloud world” where copies aren't made of data at all. Following these heuristics when providing and/or accepting data is an approach that creates at least the opportunity to be forward-compatible for the future development of technologies.
Although most scientists are likely unaware of this at present, data licenses are going to become increasingly important in science in the future, especially as we see more scientists embracing open notebook science, open science, and open-access publishing, and funding bodies promoting the increased accessibility of the fruits of their funding. We are likely not too far from funding bodies mandating immediate release of all data and results produced by each of their grantees, which is something we would advocate as potentially disruptive in its own right (S. Ekins et al., unpublished data).
We can hence imagine a near future in which many scientists will blog some or all of their research results while data aggregators will in turn consume this content and repackage it for others . The licensing of this and other data will need to be clear if we are to build on the shoulders of giants and not have to face legal battles that pit Davids versus Goliaths. Considering data licensing as a part of the “scientific process” is vital for its future usability, and we strongly encourage scientists to consider data licensing before they embark upon re-using such content in databases they construct themselves or in the course of their research.
The four simple rules we have formulated for licensing data for open drug discovery represent a proposed starting point for consideration by database producers. These licenses could equally be used by individual scientists on their blogs and other online environments or accounts in which they make their data and models available for others.
This consists of a discussion in three sections:
• Intellectual property rights in data: Copyright and Database Rights.
• Trends in legal certainty: Open Data Licensing.
• “Informal” Openness and Open License Limitations.