Evaluating and selecting the SciWfMS best suited to your lab's processes
At its heart, a workflow is simply a set of procedural rules used to coordinate tasks between people and systems, while ensuring that all steps and requirements of the process are correctly followed. Workflow management systems (WfMSs) were developed to carry out those processes.
Initial systems were relatively rigid in applying the process rules and occasionally had issues with handling exception conditions. Current state-of-the-art systems are much more flexible in handling these exceptions and in communicating with other informatics systems. However, there is no description consensus regarding what constitutes a workflow management system. On the low end, you might find some organizations using a simple spreadsheet application featuring a routing list that is supposed to be checked off as the process moves from step to step. On the high end, there are automated systems that are capable of maintaining a reliable audit trail of all activities, including who performed them and when.
Even if we discount any spreadsheet systems at the start, the intrinsic functionality of systems sold as purpose-built WfMSs is anything but standard. When evaluating systems, some people break them down into two classes: workflow and workflow light. A full-blown workflow system includes significantly more logic. It may allow you to attach multiple files to a single process, apply security controls, handle staff scheduling and consumables monitoring, etc. A workflow light system contains the most basic features and is in the main designed for document handling.
While some of these systems are designed to be very general purpose, which usually implies that more configuration is involved, some are optimized to work with a specific type of process or industry, thus requiring less configuration. This has led to the development of scientific workflow management systems (SciWfMSs).1 Superficially, a WfMS for business and a SciWfMS can appear very similar; however, they are built on considerably distinct execution models, resulting in fundamental variations between them.2
At the start of the selection process, you should perform a high-level review of the applications you are considering to ensure that they include the capabilities you require. This should be followed up with a series of online demos to narrow the field, after which you can get down to the serious business of writing up your Request for Proposal to send over to purchasing.
An implied caveat to this is that you actually know what capabilities you will require. For this to be true, you have to have a solid understanding of what your current workflow is. In my experience, most organizations have only a general understanding of their workflow. If you ask them to draw a flow chart of their process, you tend to end up with an idealization of the process; that is, what the process is when everything works properly. It is only through further examination, usually with those closer to the actual execution, that you discover all the exceptions that can occur in the workflow.
A better way of obtaining detailed information is to have the analysts in the group generate a flow chart of their processes and then have someone shadow them for a while to confirm that this actually is what they do, and discuss with them why they do things that way.
As you collect information regarding what your people actually do, you can take one of two courses of action. You can either update your standard operating procedure documents so that they accurately reflect your operations or you can take the more prudent, if potentially more time-consuming, course of using this information to reengineer your processes to make them more efficient, potentially increasing productivity and reducing the risk of generating erroneous data. While this might add significantly to the upfront cost, taking the time to do this will pay major dividends in boosting throughput, catching errors, and overall improving the quality of the data generated.
With over 200 workflow-related systems available, this screening is a critical part of the process. One way to help narrow the field is to determine which of the three basic types of workflows you are working with and use that to potentially cut out a large block of prospective systems. The defined types are:3
- Sequential workflow (while typically flow chartbased, execution progresses from one stage to the next and does not loop back)
- State machine workflow (progress from state to state, these workflows are more complex and return to a previous point, if required)
- Rules-driven workflow (implemented based on a sequential workflow; the rules dictate the progress of the workflow)
We will now narrow our focus specifically to laboratory operations. Unfortunately, there are additional questions that must be answered, as the functionality required will depend on the type of lab you operate. As a general class, SciWfMSs are designed to handle large volumes of data in multistage simulations in a scalable environment. Pegasus4 is a good example of such a system. It was first developed in 2001, and since then its capabilities have been extended significantly. A key point of Pegasus is the isolation of the workflow description from the description of the execution environment. This allows the workflow to be portable over a variety of execution environments. It also allows the system to optimize the performance of the workflow for that environment. Other workflow management and supporting systems include Galaxy, DIET, ADAMS, Askalon, Moteur, Kepler, Triana, Nimrod/K, and Makeflow.
Focusing now on chemical and clinical labs, with their respective laboratory information management systems (LIMSs) and laboratory information systems (LISs), a workflow management system has advantages to offer.
Via appropriate queries of the LIMS/LIS database, it has been possible to track the sample workload of the laboratory to aid in budgeting and personnel assignment. However, the initial LIMS products were basically hard-coded with a simple linear workflow. If the laboratory’s workflow didn’t match this embedded workflow, you were stuck with either paying for modifications to the LIMS program code or coming up with creative use of existing LIMS features.
Today’s informatics systems tend to be much more configurable regarding their behavior, but there are still systems where any major modification of behavior still requires customization of their application code. Whenever the latter occurs, best practices require you to revalidate your system. If you are working in a regulated industry, this revalidation is not a suggestion but a requirement.
There are exceptions, but many commercial LIMS/ LIS vendors appear to be moving toward product standardization, coupled with extensive configuration capabilities. While this allows you to modify multiple aspects of the system’s behavior, permitting the fixedstate workflow model to give the appearance of more flexibility in use, they are still restrictive. This is a situation where site-designed workflows allow you to control the workflow without code modifications. Fortunately, many vendors are also integrating SciWfMS features into their systems as well.
Benefits of a SciWfMS include:
- Allows the laboratory to define the rounding rules for a specific test, based on client requirement, reducing the number of test versions that must be maintained.
- Allows the laboratory to track analyst certifications for particular instruments and lock them out if their certification has expired.
- Allows the laboratory to track instrument service logs and preventive maintenance, locking the unit out of service if it is past its scheduled preventive maintenance date.
- Allows the system to schedule reflex tests when defined criteria are met. This may mean rerunning the test or scheduling an alternate confirmation test.
- Allows the system to alter test specifications based on the condition of a sample, e.g., alter the limit of detection of the test due to contamination, alter the confidence limits of the results, or switch the test method performed.
- Allows for the alteration of test completion criteria, which would be particularly useful for research laboratories.
- Allows the definition of the criteria to be used to determine the triggering of a status change or a transition between different workflows.
For an integrated LIMS/SciWfMS, the configuration of the application and the creation of the site-defined workflows should be distinct and separate,5 reducing potential confusion and simplifying both. For this type of system, a graphical environment for creating, monitoring, and troubleshooting the workflow is preferred for ease of use.
Another point to consider when selecting a system is that LIMS project teams are frequently under severe pressure to get the system installed, tested, and operational as soon as possible. To help reduce this stress, it is wise to ensure that the vendor supplies a basic default workflow framework, allowing analysts to use the system while the project team goes back to customize the site-designed workflows for each test.
Pragmatically, you must determine what data types you want the LIMS to manage, beyond samples, analyses, and results. This might include monitoring the expiration dates of standards or reagents, the analysts’ certifications, instrument service logs, etc. The system must be configured to capture sufficient data to allow the troubleshooting of all workflows and how samples were handled. This is particularly true in a regulated environment. All changes, whether of data or workflow, should be captured in an audit trail that is secure from modification.
The bottom line is that a workflow management system can enhance your laboratory’s operation, no matter what type of lab you have. Whether a contract laboratory, a clinical laboratory, or a genetic laboratory processing mountains of data, a SciWfMS, whether stand-alone or integrated into a LIMS or LIS, can improve the quality of your data, eliminate unnecessary staff operations, and boost productivity.
1. Tretter, D. Workflow engines in the different areas of application - An overview of functionalities. (Technische Universität Kaiserslautern, 2012). At http://dspace.icsy.de:12000/dspace/bitstream/123456789/357/1/20120105%20-%20Workflow%20engines%20in%20the%20different%20areas%20of%20application%28final%29.pdf
2. Migliorini, S., Gambini, M., La Rosa, M., and ter Hofstede, A. H. M. Pattern-Based Evaluation of Scientific Workflow Management Systems. 98 (2011). At http://eprints.qut.edu.au/39935/1/39935.pdf
3. PNMsoft. Workflow Tutorial - What Is a Workflow? PNMsoft (2012). At http://www.pnmsoft.com/resources/bpm-tutorial/workflow-tutorial/
4. Deelman, E. et al. Pegasus, a workflow management system for science automation. Future Gener. Comput. Syst. 46, 17–35 (2015).
5. Lococo, D. Workflow Perspectives in LIMS. Scientific Computing (2005). At http://www.scientificcomputing.com/articles/2005/10/workflow-perspectives-lims