January 10, 2012 | Inside the Box | In most organizations, the human resources of bioinformatics are a regular source of tension. Unless you’re particularly lucky, you can be plagued by politics, illogical decision making, disappointment, and low productivity. While you can have these problems in a properly-balanced organization, there are certain org charts in which they are endemic.
Before discussing the bad organizations, let’s briefly enumerate the good ones. While there may have been much groping for solutions early on, modern genome centers and large core facilities consider a dedicated software staff as a given. This full-time, in-house group works exclusively on the software and IT requirements of generating and translating large amounts of sequence data from the instrument into a usable result. Organizational questions about reporting structure, strategic direction, and customer relations are mostly non-existent due to consensus about the nature and scope of the task. The group may be part of a larger informatics organization that provides external services and software, but the team is dedicated.
A genome center can afford to have key technical teams of optimal size (between 5 and 9) that are given relatively focused tasks. These groups are not plagued with the single point of failure and turnover risks of smaller teams and can afford some intra-team specialization (e.g. UI-oriented developer).
At the other end of the scale, a single, skilled individual—more often ‘home grown’ than recruited— works directly for a lab, with either no other management, or a very weak link to a line manager. This individual can directly solve most data analysis problems through a combination of small databases, workgroup tools, analysis web sites, and downloadable software. Communication difficulties are drastically reduced because there is no development /analysis team and understanding lab personnel and processes quickly becomes internalized. Problems can arise down the road, however, since the individual rarely leaves enough documentation for his or her replacement. While these individuals will need to draw on external resources to accomplish certain tasks from time to time, with the correct mix of friends, consultants, and software vendors, the lab can usually solve a wide range of problems.
The reason these two organizational structures are successful is that many of the variables associated with bioinformatics service are set to a constant. The intra-team communication and informatician-customer communication issues are set to “1” for the dedicated individual. While the genome center support staff must communicate well internally, their customers and tasks are fairly constant. Far more common are the tense organizations where customer, task, and team are allowed to float.
Lab scientist with added bioinformatics responsibilities: In some smaller laboratories, or those venturing into higher throughput sequencing or arrays, a lab scientist with a predilection for technology can begin to take on the automation of data processing and analysis that is overrunning spreadsheets. This quickly becomes a demanding task that competes with lab responsibilities. This was far more prevalent 10-15 years ago when the idea that lab would need software/IT support was novel. The individuals that I’ve known in this category usually become unhappy because of the variability and uncertainty in day-to-day tasks. This position either decays into a dedicated informatics resource, reverts back to pure lab work when another solution is found; or the individual leaves for a pure lab or pure informatics job.
The Bioinformatics Core: In this setup, some mix of informatics analysts and developers are put together under a common organization that then distributes their time to affiliated laboratories. This makes sense in companies where line management tends to be skill-oriented and is a natural analog to the wet lab core facilities that are commonly established in companies and universities.
But this never works the way people expect it to. Informatics does not fit the core facility mold. One could imagine a disciplined core with a menu of options and strict time frames for them (R scripts for gene expression analysis: 1 month maximum; SOLiD RNA-Seq analysis: 2 month maximum). Coupled with a charge-back system, labs would make careful, focused decisions on narrowly scoped, time-limited tasks. The tasks, at least, would be set to a constant. (Un)fortunately, software development and analysis is so broad that it can be applied to a number of tasks in any given lab. Robotics, LIMS, algorithm development, specialized analysis, statistics, visualization are just of few of the varied dimensions in the bioinformatics universe. A good sized group of 5-10 can cover any of these and more. And charge-back systems are rare, so core staff members are usually “free” resources for the affiliated lab.
The fungible nature of software and analysis and the “free”-ness of the resources combine to drive decisions toward politics, informatics staff interest, and “relevance”. There is nothing inherently wrong with an economy driven by favors, but it does not scale. The lab with the largest mindshare and political weight usually gets the lion’s share of attention, leading to competition to be associated with the more important projects. Other requests get served in such a way as not to tie down resources for long periods. You can get the core to build a web application to run a custom script, but you can’t get the core to do something as general as “improve efficiency by automating critical tasks.” As priorities are vague, staff interest will start to drive decisions as well. Vaguely optimistic assignments like “prototype cloud-enabled algorithms,” will always win out over critical but unglamorous tasks like “migrate all legacy data into the new LIMS.”
A centrally-managed bioinformatics organization can function under different models than the “Core” described above, but organizational fear often prevents adoption. A core with a small menu of well defined services, charge-backs for disk and CPU usage, and a team of scientific consultants can be highly effective. Affiliated labs know well ahead of time what they can expect and what they will have to do for themselves. They also know that the consultant’s time is not free. However, as most managers can see, this is a prime outsourcing target. No matter how cost effective and customer friendly the group, some executive looking to have impact will eventually replace this organization with an external company.
The loosely coupled local expert model where the “skilled, dedicated individual” described above is part of a larger bioinformatics organization can work. However it assumes that the bioinformatics management is pretty hands off, doing little more than communicating best practices among the group helping to screen new applicants. Without a capable leader, though, the temptation to absorb the dedicated resources into the lab is very strong. Again, we decay into the “skilled, dedicated individual” system.
Informatics support is an increasingly necessary part of laboratories of all sizes. The pace of change has so far prevented the establishment of shrink-wrapped solutions and necessitated the application of the immensely powerful human brain. Getting the organizational structure around that brain right can be the difference between years of productivity and constant politics, reorganization and lost value. •
Aaron Kitzmiller is a senior scientific consultant with the BioTeam. He can be reached at firstname.lastname@example.org .
This article also appeared in the January 2012 issue of Bio-IT World magazine.