Adjusting a Linear Based Calibration Curve

Scientits to Scientist

Instrumentation

Automated liquid handlers (ALH) like the Beckman Coulter Biomek are great for improving productivity and reproducibility in the lab, but you can get even better performance—and correspondingly more reliable data—when you optimize the instrument for accurate liquid transfer.

In this video, we are using a Beckman Coulter Biomek NXp 96-tip ALH and the Artel MVS Multichannel Verification System. However, this optimization process can easily be used with other liquid handlers, such as the Tecan EVO, Agilent Bravo, PerkinElmer Janus – and others that use a linear-based calibration curve (as in y = mx + b).

Calculation

In order to optimize performance, many liquid handlers allow users to adjust the calibration curve to compensate for differences between expected and observed volume. By plotting the expected volumes (y) vs. the observed measured volumes (x), a linear regression allows a new slope and offset value to be calculated.

Note: The y = mx + b equation may vary depending on the automated liquid handler manufacturer, make and model.

The initial steps of the optimization process work like this:

  • Test
  • Measure
  • Assess

Test

Start by testing. You want to test the three target volumes that span the desired volume range. In the example depicted in this video, a simple method was scripted on a 96-tip Biomek NxP for testing three target volumes between 2 & 8 uL with the calibration curve’s slope and offset values set at 1 and 0, respectively.

For this method we used new P20 disposable tips for each single transfer of an Artel dye at 2, 5, and 8 uL into a Diluent-filled 96w microplate.

Measure

Next, measure the as-found volume transfers with the Artel MVS. After the liquid transfers are completed, measure the three test plates on the MVS – which will take less than 10 minutes.

Assess

As you can see from the low CV values below in the table, the data are repeatable for each target volume, which is a good thing – the accuracy, however, is low by 7-19% over that small volume range.


Because of the low CV values, the transfer accuracy over the volume range can be improved upon as long as there is linearity between desired target volumes versus measured target volumes.

The average measured volumes from the MVS test reports, as well as the original slope factor and offset value, are plugged into the linear regression calculator. In the image below, we show how our in-house calculator was used to determine linearity over that specific volume range.

Check for linearity. In this case, the results are very linear as noted by the R-squared value of ONE.

Because it is linear, we can use the newly generated slope factor and offset value to adjust the liquid handler software and repeat the process of test, measure and assess. Before retesting the liquid handler, the new equation’s slope (1.035) and offset (0.318) are entered into the liquid handler software.

Using the same scripted method, except now with the new slope factor and offset value, three new test plates were dispensed and again measured with the MVS. The “post-optimization” test results shown below drastically improved where the accuracy over the volume range is now 0.4 to 2.3%. This graph shows the measurement data, before and after we adjusted the liquid handler’s calibration curve.

You can now have confidence using this optimized method over that volume range. As always – if liquid types or labware are changed or replaced, this optimization process may have to be repeated.

Working with well-defined liquid classes can cut the time it takes to get an assay into production and improve assay quality by reducing variability stemming from liquid handling.