For Merck, scientific computing isn’t some ivory-tower exercise. Genetics, biology and medicine (an applied science) are all fundamental to Merck’s business. But scientific computing poses a different set of challenges than those that arise in everyday business applications. The biological and medical domains feature complicated taxonomies, millions of technical terms and an ever-changing terminology.
Problem: Scientific terms change radically and often.
Resolution: Monitor changing terminology and update databases frequently.
“There are so many names for things, and they’re constantly in evolution,” says Stan Kaufman, a cardiologist turned software developer who consults on clinical research through his firm, The Epimetics Group. “When you think you’ve finally got a handle on it, something new gets discovered and you find you have a whole new set of terms to deal with.”
In addition, there are often multiple terms for the same item. “A given biological process will get called entirely different things depending on whether the doctor was trained on the East Coast or the West Coast,” he says.
So far, attempts to create universal terminology standards or automate the translation between different terminologies have met with limited success, Kaufman says. So scientific computing site managers must be willing and able to evolve databases and applications to keep up.
Problem: Doctors and scientists often find that the records they want to keep don’t fit into the neat confines of a data entry form.
Resolution: Address legitimate concerns, but impose the formats, fields and structure of a good database. Take the example of clinical testing of a drug. For the developer, the ideal user interface employs check boxes or drop-down lists that restrict the doctor to choosing from a pre-defined list of common side effects. That way, when the data is analyzed later, the application will be able to tally how often each side effect occurred.
Problem: Not every item a doctor needs to record is covered by the system.
Resolution: Find ways to grab “free form” information—and then let users find it.
When filling out paper forms, doctors like to check a couple of boxes, then scribble notes in the margins. They tend to want to do the same with computer systems—which don’t accommodate writing in margins.
Terrence Critchlow, a data-management project leader at Lawrence Livermore National Laboratory, runs into the same issue working with genetics researchers. His compromise is to accept free text entries and then try to “identify and pull out the important information as best we can” by identifying keywords in the text.
Problem: Scientists may not want to conform to the standards that apply to other business units.
Resolution: Accommodate the need—if it creates a competitive advantage. Make sure that the request is not just a personal preference. Determine if they have a real need that can’t be met by the standard platform.
If, for example, a researcher’s groundbreaking work hinges on access to some piece of software that only runs on an obscure variant of Linux, Merck CIO Chris Scalet says he would “have no intention of going in and squishing that innovation—that would be suicidal.”
