This post follows on from a conversation with Christoph Borchers of the University of Vancouver in which he states “Accuracy is not important, precision is everything”. This is a highly provocative statement, especially in the analytical world but is he right?

To begin with I have to introduce some key analytical terminology:

Accuracy – how close the calculated result is to the true result

Precision – how close replicate measurements are to each other

Ideally you want to be accurate and precise but can also be accurate but not precise, precise but not accurate or even not accurate or precise.  

So how do you know which one of the possibilities describes your analysis? You can easily measure precision but to measure accuracy you have to know what the true value is and this is very difficult when dealing with proteins. Proteins are complicated large molecules and even creating standards is tricky, never mind putting them in ‘blank’ matrices (the analytical term for the environment your protein would normally be in e.g. plasma or a whole cell lysate) to test your assay.

If accuracy is so difficult to measure then what relevance does it have? The question here relates to how you apply your quantification. If the measurement is made to determine whether the levels are with a ‘normal’ range or outside it then what is really required is that the same sample gives the same results over an extended period of time. This now brings in two uses of the concept of precision:

Intra-batch precision – all results within the same batch of samples agree with each other

Inter-batch precision – samples analysed in different batches on different days agree with each other, this can also be termed repeatability or reproducibility

It can therefore be seen that actually it is inter-batch precision or repeatability that is the most important. To achieve this you need highly reproducible sample preparation along with robust analytical equipment with systems for tracking variability in performance and correcting for it. This means you need multiplexed methodologies for all stages and preferably robotic automation for sample prep. The mass spec analysis has to be comprehensive and have standards included for correction. Recent presentations by Jennifer Van Eyk (Cedars_Sinai) and Tony Wheaton (Manchester) give some examples of the future and it is big and automated!

So in summary I’d alter the initial statement somewhat to give:

“Accuracy is often not as important as inter-batch precision/reproducibility”

Tags: #facility, precision, accuracy

Related