It’s roughly 10 years since the majority of large biopharma companies selected and deployed an electronic lab notebook, ELN, to their research group for Experiment Capture workflows. The early adopters were the chemistry community, driven by their need to better protect IP and improve the efficiency of their experiment write ups. Subsequent to the chemistry capability, biology and especially bioscience and DMPK/ADME were the next groups to make the transition from paper notebooks to ELNs. Smaller biotechs and the wider research community have since followed suit and adopted ELNs helped by the more recent ability to use Cloud-based (SaaS) ELN deployments. These have made the adoption process somewhat simpler. There is still, however, a large community of for instance academic research organisations that continue to use paper notebooks and are yet to see the fundamental value of ELNs.
However, the biopharma industry has changed hugely since the mid-2000s not just in the continued growth of biotechs looking to exploit new target opportunities and driving innovation for new modalities, but equally in the desire for more data and better data re-use to feed the expectant and data hungry AI (Artificial Intelligence) and ML (machine learning) pipelines.
We are now at a point where those 10 year-old and older systems (primarily larger thick client deployments) are facing a number of pressures within biopharma. These include, for example:
- Lack of flexibility of the thick client and the installed platforms;
- Age of the 1st wave thick client systems, and their cost and effort of maintenance. This includes ever-increasing time taken for upgrades driven by the fact that many have had considerable customisation beyond simple configuration;
- Lack of support for biopharma collaboration models using on-premise installs;
- Patchy support for digital lab and the growth in data volumes and diversity;
- Poor fit for the new science domains especially molecular biology and also data science workflows
- Ability or to be more precise, inability to extract data (we cover more on data below).
More importantly, there is a far larger issue around whether these ELNs have supported appropriate experiment data capture in formats that allow both effective primary analysis as well as re-use as part of AI/ML and longer-term machine based analysis.
All too often, ELNs have become mere dumping grounds for data including raw and processed instrument data (“refined” data) with little or no metadata. Without good metadata or tags, these files are effectively lost forever in the ELN systems.
In the beginning, experiment capture was enabled, to an extent, by ELNs, but more recently this capability has not been supported effectively; for example, the details of the experiment are often in unstructured text or poorly annotated additional files. This renders their usage limited and low value.
Now seems a good time to reflect on what ELNs have actually helped with, and also, more critically, to help plan the next phase of experiment data capture so that the voracious animal that is AI/ML can be more effectively fed.
It should be stressed, though, that the future of appropriate, value-added experiment capture is not just about ELNs. Consideration of the wider ecosystem of tools is needed, as well as the business change that needs to happen to support the vital aspects of improved metadata capture and early potential of FAIR¹ (Findable, Accessible, Interoperable and Reproducible) support and projects such as FAIRplus.²
2019 feels like the right time not only to review where we have got to with ELNs, but also to consider support for a more holistic approach to experiment capture since that is value adding activity not the ELN tool itself.
Experiment Capture Workflows Review
As part of that review, you may wish to consider some of these questions as you look at your existing ELN deployments and experiment capture workflows.
- With regard to the effectiveness of your organization’s ELN(s) in capturing key experiment information, what data are you missing out on?
- Have you considered the data lifecycle beyond the ELN and your wider ecosystem?
- What effort is required to support the ELN and its adaptability?
- Do you have the correct ELN(s) to support new science and their capabilities (e.g. molecular biology and new modalities)?
- Domain specific ELNs can be valuable in enabling key science workflows
- How does your ELN support collaborative science? Can it effectively handle different models of collaboration across organisations?
- How does your ELN handle experiment metadata, bearing in mind the critical need for better-defined data and metadata to drive ever more important automated analysis (AI/ML and Data science)?
We look forward to these ideas being discussed in the coming meetings and events such as Paperless Lab 2019 and BioIT 2019
Note: The author is involved in IMI FAIRplus² project
References:
- FAIR data principles https://www.force11.org/group/fairgroup/fairprinciples
- FAIRplus https://fairplus-project.eu/
- Photos courtesy of Unsplash
- Photo by Drew Hays on Unsplash
- Photo by Katya Austin on Unsplash