Guidelines for submitting manuscripts for publication require you to consider both technical and biological replicates. A typical example can be seen in the notes to authors from PROTEOMICS and PROTEOMICS – Clinical Applications (Wiley-VCH).
To satisfy this, many people run both biological and technical replicates as part of the same experiment. This adds unnecessary complexity to your data analysis and, unless handled correctly, you risk pseudo-replication.
As an alternative, you can consider setting up experiments that comprise multiple technical replicates of one biological sample. These are then used as a measure of the technical variation of your system and approach. You then compare any changes observed between biological replicates against the background of technical variation to report the reliability of the biological differences.
In this post, we’ll consider an experiment where both types of replicate are run together, since this is often what occurs due to limitations in time or resources. Defining your experimental groups as containing both technical and biological replicates would introduce bias and error, making your results invalid.
So, how do you analyse both technical and biological replicates to give statistically valid results?
Analyse technical and biological replicates within a single experiment
The approach to analysing a mixture of technical and biological replicates with Progenesis is like reproducing the whole experiment, but doing this easily in silico, and comparing two or more sets of results to find significant features common to both. This is true whether you use Progenesis QI for proteomics or Progenesis QI.
Summary: Progenesis products can be used to analyse experiments that contain a mixture of technical and biological replicates in an organised, statistically valid way. The result is like reproducing the whole experiment and comparing two or more sets of results to find significant common to both, the intersection in the Venn diagram above.
How do you do this in Progenesis?
The key is in the unique way Progenesis products align and co-detect runs, which means the same features are detected and measured identically on every run. This allows you to easily compare different views of the same data, comparing features across any runs that are set up in different experimental groups. We can demonstrate this with an example of a study comparing Control versus Treated subjects containing a mixture of technical replicates, A and B, as well as biological replicates 1, 2 and 3.
- Figure 1: Set up all files in one experiment, align and peak-pick then set up multiple experiment designs within the same experiment. One experiment design should contain all files together and the other experiment designs only contain the technical replicates of runs from each condition. In our example below, we have created an experiment design with all technical and biological replicates from the control subjects versus the same for Treated subjects. We also generate an experiment design comparing technical replicate A’s for Control versus Treated and another comparing technical replicate B’s for Control versus Treated.
Figure 1: Set up all files in one experiment, align and detect then set-up multiple experiment designs in the same experiment.
- Figure 2: Once you step into the Review Peak Picking section, you need to: 1. Select the experiment design based on containing technical replicates, A or B in our case, from the drop down option (highlighted in red below). 2. Apply tags to features considered significant based on your chosen criteria within each experiment design. Then apply tag filters to view the discoveries common to both replicate analyses, as well as those that are common to only one replicate set.
Figure 2: 1.Tag features with significant criteria e.g. p-value <0.05, fold-change >2 of Control versus Treated in both experiment designs i.e. All technical replicate A’s for control vs. Treated and all technical replicate B’s for Control vs. Treated. 2. Apply tag filters to only list the features with significant criteria common to both experiment designs.
Figure 3: By using the tags and applying tag filters, you can identify the specific features that are common to each analysis. From here, you check the number matches your FDR and these features can be ignored in any necessary validation work.
Figure 3: The final analysis results from the example of Control versus Treated samples made up of a mixture of technical and biological replicates. Here we could take the 68 common significant features and select those for further study.
Want to learn more?
This is a very simple overview and example, so you may wish to see a demonstration of this with your own data. Get in touch, and we can arrange to set this up with you. Hopefully we can also show you what else Progenesis QI for proteomics or Progenesis QI can do to help your proteomics or metabolomics analysis.