Choosing the right Parallel Line Analysis (PLA) software for your lab is not an easy task. Whilst there are several packages available, and most can perform an adequate standard job, at least for development, it can be a challenge to know what aspects to consider. This is an important consideration as choosing the right software contributes significantly to product and business success.
Generally, relative potency bioassays are both biologically and analytically complex. Early progress on the analysis of data usually begins with spreadsheets, but soon you will want to try fitting complex models such as the 4- or 5- parameter logistic (4PL or 5PL). Dedicated software will be required. This dedicated software performs the complex Parallel Line Analysis (PLA[1]) with the necessary security for GMP production.
(For more fundamental knowledge of relative potency please see our blog “What is Relative Potency”.)
When comparing and weighing up the features of different PLA software, the difficulty lies in the details. These details determine whether you come to love the software because it ticks all your boxes or whether you must learn to accept and live with a software that imposes scientific and business limitations. It is important to get it right, as switching between PLA software later can be disruptive and may have regulatory implications.
So, here is a short list of things to consider before you commit to a system you might have to live with for decades:
- Statistical Options in General
- Regulatory Status
- Regulatory Validation and QC
- Decide on software deployment: desktop vs local network vs cloud
- “Seat” User licence vs Concurrent User based licence.
- Statistical expertise and good support
- Upgrades and releases
- Training required
1) Statistical options in general
Do you want software that is designed for relative potency analysis and perhaps ELISA-type standard curve analysis, or do you want lots of other functions as well (kinetics, spectral analysis, etc.)?
Software with lots of functions tend to be more complex to use and require a higher level of statistical understanding.
Statistics for relative potency and ELISA-type standard curve analysis
Typically, this will include linear, 4PL and 5PL constrained models, probit and logit if you are involved in binary assays. Additionally, there are the associated data transformations, outlier analysis, and all the standard system and sample suitability criteria that regulators might require.
The key issues:
Mathematical reliability.
The maths of 4PL and 5PL PLA is complex, and calculating the right answer is not simple. In PLA analysis there is a single right answer, but some available software does not find it reliably. To ensure easy deployment across your company sites, you will want software that always finds the correct solution, regardless of software version number and operating system version used.
Ease of use (from a statistical point of view)
If you don’t really understand all the maths, it is very easy to get it wrong if you are faced with lots of options, or worse, have to enter formulae.
If the statistics is a bit of a blur for you or your QA/QC team, opt for a system designed for scientists. A software which has all the required options available from simple, appropriate selections and no formulae or unintelligible options is much easier for everyone.
If you do understand the statistics and have unusual complex assays (statistically speaking!) then you may want to buy a package with many programming options for advanced and difficult situations. (Although if you were a statistician, you would probably just use a statistical package such as R or SAS and do the programming yourself, assuming you have someone with similar skills available to check your work.)
2) Regulatory status
Do you need software for development work only, or for GMP use at a later date?
If the software is to be used for development and GMP, then look for a system that integrates these modes well. Ensure that you have access to version control so that the changes required (for e.g. reference bridging ) can be explored in a development mode without impacting the GMP mode until the change is approved.
If you require PLA software for GMP use, you may need other standards such as 21 CFR part 11. You will also need a host of other features such as audit trails, access controls, data integrity control and trending functions.
3) Regulatory validation and QC
Simply running a locked analysis on a new data set in GMP use is generally quite quick; what differentiates the software products available is the QA time associated with installation, system validation and particularly configuration validation and QA/QC on every analysis. The more the software allows you to do, the longer it will take to validate an analysis process you have set-up before you can use it at all in GMP, and the longer it will take to QC results on every analysis. If you have entered your own formulae, this process may become a major process pinch point.
How will it be done? How often will it need to be done? Is there anyone in your organisation who is statistically qualified to check formulae etc? How will they know the result is correct? Note, if you simply use test data and results created on the vendor’s version then at best you are showing it works on that data provided but at worst you may simply be ensuring that you get repeatable, wrong results.
Once a configuration is in use, QA and QC will be required to try to ensure that the system worked on that day’s particular data set. How? Is there anyone in your organisation who is statistically qualified to check formulae etc?
Think about data import and export. If it is manual, this will also have to be checked. What data format is needed? Does this require another manual step to get it in the right format? Any part of the system that is not fully locked and automated will have to be checked.
This is a big topic, and we are planning a blog about it soon. If you want to know how our software QuBAS uses Continuous real-time Validation (CrtV™) to reduce all this to almost nothing, click here or contact us.
4) Decide on software deployment: desktop vs local network vs cloud
Each solution has pros and cons, and without getting into deep IT tech-speak the basics are:
Desktop based
This is generally old technology, simple, but normally very restrictive. The software is loaded on a PC and is only accessible if you are in front of it (unless remote desktop services can be arranged). It may be very dependent on the version of Windows you are using.
Desktop based software is hard for IT services to maintain. It may not interface well with company security systems and user identification, and for GMP use may require a dedicated PC. Plus, software designed to run on a particular version of Windows may not function correctly if Windows updates itself.
Choosing a PLA system that only works with your plate reader will create larger switching costs as time goes on and compatibility issues as your lab evolves and new instruments may be introduced. Although these software systems can sometimes be used with other data, this is usually a cumbersome process, often requiring QC-heavy cut-and-paste data import.
You may also want to integrate the system with LIMS – check the software can automatically export into your LIMS.
Think about data import and export. If it is manual, this will also have to be checked. What data format is needed? Does this require another manual step to get it in the right format? Any part of the system that is not fully locked and automated will have to be checked.
This is a big topic, and we are planning a blog about it soon. If you want to know how our software QuBAS uses Continuous real-time Validation (CrtV™) to reduce all this to almost nothing, click here or contact us.
Local Server-client systems
In these systems the software is loaded onto your company’s server, and access can be gained from any computer that has access to the server. This could be via the local network, or remotely over a “Virtual Private Network” (VPN). You are not restricted to the lab or building or even a city or country.
Some older desktop software can be made to work in this way but it is not efficient from an IT point of view, and often requires the server to “pretend” to be lots of separate desktops so that remote access is possible, and more than one user can be supported at one time. More modern software designed for this use acts as an “app”, like a local website and allows access via a simple browser such as Chrome, Firefox or Edge.
Cloud services
If you move the server loaded with the modern “app” type software, out of your company’s network and put it onto the web, you have the systems we are all familiar with for banking, shopping, etc. Typically, these services would run on a server owned / maintained by the vendor rather than your company, ensuring rapid deployment, and support without relying on stretched local IT support. The issue here is to ensure security and decide how much control you want over upgrades etc.
There are a bewildering range of cloud options. If you are interested in these options, choose a system that already has a suitable structure even if not advertised as Software as a Service (SaaS) only.
Remember to think about what happens to your data if you decide to switch later!
5) “Seat” User licence vs Concurrent User based licence.
The licensing structure determines a major part of the costs. We hear lots of complaints about licensing from users of PLA software solutions. Most PLA software rely on each user having their own licence. This causes lots of admin issues as staff arrive and leave or leads to sharing of credentials – a GMP major finding!
We feel the fairest approach is a concurrent user-based licence when you can have as many users as you like. The license controls how many are active and/or using the software at the same time. depending upon the number of licences you feel you need. This also makes adding and changing personnel very easy without having to refer to the vendor.
6) Statistical expertise and good support
Some PLA software platforms are designed by software companies or instrument manufacturers. Such companies may lack staff with formal mathematical / statistical qualifications.
Bioassay analysis is complex – during your due diligence you should expect to find at least Masters level but preferably PhD statisticians involved with the design team. If not, you will need to be convinced that staff without formal training are competent to design bioassay statistical packages (see also upgrades below).
Also check what social media has to say about support services – some reports are terrible!
Statistical knowledge is important. Quantics staff have identified significant statistical issues in USP and PhEur guidelines (new guidelines have now been issued). In some well-known commercial PLA software packages that might cause assays to pass when they should fail and fail when they should pass.
7) Upgrades and releases
Most software is subject to continuous development with new functions, but new versions are also released to fix errors that come to light in use. You should check the release notes to see how often fixes are released. Some systems are releasing 5-7 times per year. Clearly this would be seriously disruptive, and you should investigate the consequences of not always upgrading. This may be fine if it is just new functions, but if the upgrades contain significant bug fixes this should warn of poor pre-release testing and control.
If the fixes are in the mathematics, it may suggest that statistical knowledge is inadequate. Mathematical bug fixes may prevent backward compatibility.
8) Training required
If the vendor offers extensive training support, ask yourself why this is necessary. The obvious conclusion is that the software is complex to use. Good software should be intuitive and easy to use and not need a 500-page manual.
Of course, adequate training should be available, and supplied free with the system.
If you decide to go with a system that needs extensive training, remember to factor in the real cost of this in money and time.
And finally
We hope you found this blog helpful and as we know looking for a suitable PLA software can be a timely process. Please do your research carefully and if there are any further questions we’d be happy for you to ask us how we handle all the above points.
As a last admission, we must apologise that while we have tried to write this in an unbiased way this is inherently impossible for us. We have spent the last 8 years passionately designing and improving our PLA software, QuBAS, with all these issues at the front of our mind. We think it’s the best but don’t take our word for it – take someone else’s!. We have some fantastic testimonials that we would love to share with you.
We (of course) hope you will come to the same conclusion as us and our clients.
Book a demo with us today.
If you’d like a draft User Requirement Specification (URS) you can use for your own business case, click here.
[1] Note: This blog uses the acronym PLA to refer to parallel line analysis, by which we mean any analysis where the test and reference dose response plots are the same shape, just shifted on the x axis. Not to be confused with PLA software by Stegmann Systems.
Comments are closed.