Curve fitting is used as the final data analysis step in many plate-based in vitro experiments and is therefore performed several times per week in most research organisations.
Yet, many labs still struggle to find an effective process for curve fitting. In this article, I will highlight the different categories of curve fitting tools and their pros/cons to point out what you need to consider when looking for a new curve fitting tool.
In biopharma research, “curve fitting” is the process of turning a set of concentration - response (or dose - response) results into a sigmoidal curve from where one can read an IC50 (ED50).
Lab scientists need a software tool implementing the relevant algorithm to perform the curve fitting. There are several different types of curve fitting tools servicing different needs.
With the relevant tool the scientists in the labs can set up a standardised process, making it easy and fail-safe to perform the analysis, while at the same time ensuring the curve fitting is done in the same way every time across experiments. The latter is a prerequisite for future cross-experiment comparison.
In general, I believe we can group the currently available tools into the following 4 categories each with their own set of pros and cons.
This is naturally not a tool as such! But all biopharma organisations have data scientists today and they - if working in research - will likely be working with Python or R every day. It will therefore be fairly easy for them to write a script to perform the curve fitting.
Pro:
Super flexible (if the data scientist has time!) and likely even tailored specifically to your environment, data, and experiment(s).
Con:
Does everybody use the same script? Is it easy to use for all? Can the script handle changes to the experiment setup? In other words: How flexible is it? Is the data scientist also available for general maintenance/updating of the script?
And last - but certainly not least - it only covers the actual curve fitting. The script does not help in the process and data handling from reader file (raw data), to assigning plate layouts and concentrations, to performing the data normalisations. All steps needed to be completed before the actual curve fit. This will therefore still have to be done manually by the scientists.
It is possible to find other standard tools - like plugins to Excel or standalone plotting tools - that can perform the actual curve fitting.
Pro:
Likely easier to understand and use for non data scientists than Python and R. Often have a lot of features related to the fitting - different algorithms etc - and the display of the curve.
Con:
Some of them are expert tools and hence require training to master. Also often have far more algorithms and detailed features than necessary for the standard use case => fit a sigmoidal curve.
But the main issue here is - similar to above - that the process up to the curve fitting is not supported. Hence, keeping track of the plates, the layouts, what compound is where and the normalisation to the controls will need to be handled by the scientists elsewhere.
By dedicated I mean a tool focused on the end-to-end curve fitting process and hence developed to handle the general process from reader file to curve.
Pro:
It handles the entire process. No integration to plate management or pre-registration of plates needed. Focused on the standard case making it easy to use. Enables a standardised process across labs.
Con:
Doesn't have a lot of features and algorithms for creating and manipulating the curve. Not a data scientist tool. Not (as default) integrated with other tools up- and downstream.
Some of the big established enterprise vendors naturally also have tools for curve fitting.
Pro:
Have every feature under the sun. Integrated both up- and downstream and hence enabling full end-to-end tracking from compound to plate to curve fit to data warehouse, as is often a requirement in big pharma.
Con:
Far more features than most labs will need for their standard work. Very flexible - but requires experts and templates. Due to the integrated nature often requires all plates to be pre-registered in an enterprise logistic solution. Complicated to implement and often not seen as "easy to use".
Before buying a new curve fitting tool it is, as always, a good idea to think about (and agree!) what actual issue or use case(s) you are looking to solve.
Is it a one off experiment? Is it complicated data and fits that need to be handled differently every time?
Then maybe find a data scientist to work with.
Is it standard plate-based experiments you need to support? Where the need is a quick and easy-to-use process aligned across different labs and departments to minimise the risk of failure and ensure the data can be compared.
Then look for a focused curve fit process tool.
Are you in the market for an end-to-end integrated enterprisy solution?
Then you know where to go!
But in general remember to consider:
At grit42, we have decided to focus on the standard dedicated process tool and have developed a stand-alone tool (gritCurveFit) that handles the entire process from reader file to curve in a simple, yet flexible manner.
Please contact us if you would like a free trial of our gritCurveFit to see how it works.