Put down that pen and paper
Paper is dead ... again. Or so say software companies. This time, they are writing the obituary of the hand-written research notebook, the demise of which was prematurely reported a decade ago.
The drawbacks of notebooks have been rehearsed many times. They have only got worse as regulatory requirements have tightened and bulky analytical data has multiplied. Put with that the age-old problems of incomplete or illegible data (due to haste or inertia), loss or damage, lack of index or search capability, and you have a compelling argument for change. But hand-written notebooks linger on for many reasons: they're easy to use; portable; and are important for legal purposes related to patents. Oh, and because attempts at electronic notebooks haven't so far succeeded in delivering what scientists actually want.
For some time now, the race has been on to improve the efficiency of drug discovery. Large pharmaceutical companies have come under increasing economic pressure and need to increase the productivity of their research efforts. The arguments have been rehearsed many times: increased pricing pressure from governments and generic competition; soaring research costs; and fewer new drugs. Senior managers agree that more experiments are not the answer - what's needed are better experiments, based on knowledge and analysis of existing information. And that requires a better notebook, an electronic one.
A new period of optimism is upon us, as existing software matures and as academics and commercial software firms look for better ways to capture, share and re-use scientific knowledge. In the UK, the publicly funded university projects Combechem and eBank aim to dissolve the boundary between laboratory notebooks, journals, the library, and teaching resources. Describing his group's work to create a smart laboratory as part of Combechem, University of Southampton researcher Dr Jeremy Frey said: 'This is more than just an electronic notebook, because those have sort of come and gone and not been ideal.'
Previous attempts at an electronic notebook sometimes failed because they didn't look closely enough at how scientists actually worked. Perhaps the technology wasn't available to support the kinds of, or level of, cooperation or searchability that individuals expected. Scientists were generally unimpressed; Trevor Heritage of informatics software firm Tripos suggested. 'I have 10 or 15 different tools on my desktop that I have to use each day to get my job done. So you've just given me the sixteenth. Thank you.' Perhaps they lacked flexibility, the ability to integrate easily with other software, or were difficult to scale up for widespread use throughout a large multinational company. Not least, few were designed for data-mining by non-specialists.
Like the Combechem project and unlike some previous notebook software, Tripos' Electronic Notebook started with the working practices of laboratory scientists, rather than the workings of analytical instruments, for example. It began as an internal project at Tripos' chemistry labs, which produce drug targets for pharmaceutical companies. On a green-field site in Bude, Cornwall, Tripos was able to create the operation from scratch. 'We made the strategic decision that we would be a process-oriented chemistry company, and that we would provide the informatics software to facilitate those processes,' said Heritage. 'We imagined that the concept would be of interest to our customers when we started to build the technology for our own operation,' he added. But he hadn't expected it to become a product - the Electronic Notebook.
The change occurred because, over time, pharmaceutical and biotech companies also became convinced that informatics offered a way of improving the efficiency of their research arms. 'They have themselves become more process-oriented and less serendipitous in their activities,' said Heritage. Companies that visited Tripos Discovery Research for help in their research programmes saw the benefits. 'It [the software] enabled our scientists to work and learn from previous mistakes,' said Heritage. And some then became interested in having the same kind of system.
The compelling evidence for its utility came from the greatly reduced time needed to synthesise analogues of successful drug targets. Tripos ships compounds, made under contract, to its customers who screen them for 'hits' (signs of biological activity). 'Typically, once they get hits, their next question is, "we'd like some analogues",' said Heritage. A typical turnaround from their own scientists or a competing supplier is six to eight weeks, maybe more; for Tripos, it's two weeks. The difference lies in the informatics software. If a customer comes back with a request for similar compounds, 'we can go to the Chemcore system [which links to the notebook], give it that list, and say find 100 compounds that meet these criteria and can be synthesised in this likely yield'. Typically Tripos' customers don't know which reactions are likely to fail and they end up running many more of them, said Heritage.
Then the notebook secured its first customer - Schering. At the time, Schering was considering outsourcing a chemistry project to Tripos. After visiting the labs, the company asked Tripos to implement an electronic notebook as part of a similar system for them, and an agreement was signed in September 2001.
Tripos first set about understanding Schering's existing systems and workflows, and then reconfigured the software to work outside its lab. This is a difficult step, said Heritage, and it's where many companies fail. 'They develop something really cool for themselves, but can't export it.' But Tripos had its own software consulting division with just the sort of experience to get the job done. 'We identified parts of the system that are reusable and exportable, and then re-implemented the architecture so that it's more product-oriented. And we are currently rolling this out with Schering.'
That architecture is Java-based, with a client linked to an Oracle database through an application server. Notebook pages are stored in an Oracle database. This can be coupled with Tripos' chemical data cartridge, Auspyx, which can store and search chemical structures. The use of an Oracle database ensures that data can be accessed through standard technologies and by future knowledge-management applications, should a user adopt software from another supplier.
The middle tier uses Java application servers, such as BEA's WebLogic, or the open source alternative JBoss. As well as communicating with the Oracle Database, this tier can also link to laboratory information management systems (LIMS), inventory and document management systems, as well as any other existing databases that contain important information. It is also where business rules are defined and executed.
The Java client runs on a Windows 2000/XP workstation. A notebook is made up of a collection of pages, with each page assembled from components representing types of data or pieces of workflow, e.g. reactions, spectral data, literature, and references.
A new experiment starts with a template notebook page. The components of the page are set by company or department rules, but individual scientists can add things or move them to suit their individual working practice. Only the author can enter information and alter pages. But access can be granted to other users or groups through settings made by the system administrator, who can also add new page components to the existing notebook as and when needed. Which raises the question of privacy.
Handwritten notebooks aren't just convenient to write in and to carry around; they retain an element of the personal. In Tripos' notebook, privacy is handled in a variety of ways. Most scientists don't want to share an incomplete work. 'So it's only published after they hit the finalise button for that experiment'. Beyond that: 'The system is entirely configurable,' said Heritage. It is also possible to create some private fields, which are still stored within the database but are not made accessible to anybody else that retrieves the page.
There is also a planning page, which is private. It can hold ideas, references, reaction schemes, and other notes. It is stored inside the database, and is attached to an experiment when it is eventually carried out, but only the scientist that created it can see it.
The dates on which pages are created or modified are logged, and once a page is completed it can no longer be changed. (If a scientist needs to modify a finalised page, this can be done by creating an amended page along with the reason for the changes.) The page is then printed and sent to a document management system, where it is locked and time-stamped. The hard copy can be signed by appropriate managers and stored for legal purposes. The date an invention was first conceived and the date it was first put into practice (e.g. the first chemical synthesis) are both potentially important for 'first to invent' patent claims. If it seems odd that, after all these attempts at data integration, the record ends up on paper, put it down to the conservatism of the legal profession. 'The pharmaceutical industry will continue to print these things out, up until the point in time where there's a successful patent defence based purely on electronic mechanisms,' said Heritage. In any case, a lot of companies are really looking at the notebook for the workflow optimisation, which is where they see efficiency savings coming from, he added.
Tripos is now preparing to roll out the notebook to more than 500 scientists in Schering's labs around the world. Beyond that, it won't talk about any other customers. 'We've only recently got to the point where we have this combined offering of the service-oriented analysis and the technology backbone to support it,' said Heritage. Nevertheless, Tripos is already working with companies to help them understand their workflows. This part of a consulting project takes around three months and, at the end, they are handed a document that shows their existing workflows and their needs. 'Then we jointly identify ways of filling those gaps.'
'Successfully deploying an electronic notebook, or in fact any workflow-oriented system, is much harder than just installing a piece of software on a desktop of the scientist who's going to use it,' he said. A fairly deep understanding is needed of all the processes that scientist is going to use and all the systems that he or she depends on to get their work completed, argued Heritage. 'As you get deeper into it, you find the customers don't know their processes as well as they think they do'.
Optimising workflow Typically, Tripos finds that scientists have adapted the way they work to fit their environment. But following those original steps won't work when installing a workflow-optimisation system. 'Scientists know that's a sub-optimal process, but it's the one they were forced to use,' Heritage believes. By looking at what causes a scientist to pursue a certain set of operations, and what he or she is trying to accomplish, can lead to unnecessary operations. 'We are not proposing, or changing, scientific processes for the joy of it. If it doesn't need to be done, we implement the interface so it supports the existing workflow.' This rigorous approach should make resistance to new ways of working less likely, he believes.
Because Tripos is a consulting firm, as well as a software supplier, it would like its customers to understand why they want an electronic notebook and what they want to achieve with it. 'Everybody thinks they want a notebook, because they hear that other people want a notebook,' said Heritage. But few organisations have created a business case and the budget to implement and deploy a notebook on a scale that will realise the full benefits. 'Our goal is to help them to understand why they're buying this thing, what workflows it's going support, and how. And what the expected benefits, in terms of workflow and return on investment, will be.'
So a new approach to electronic notebooks could satisfy scientists and pharma company managers alike. And nobody else is taking quite the same approach as Tripos, claimed Heritage. 'Probably the closest company doing something similar to us is IBM, [but] doing it from a knowledge-management and consulting prospective'. Which poses the question. Can Tripos' emphasis on workflow, process optimisation, and knowledge-management, offer something to companies beyond its niche in drug discovery? 'I think it's widely applicable,' he said. So watch out, IBM.