Replicates
Performing repeated measurements from multiple study subjects to assess biological and technical variability within a study.
Click here to download a fact sheet on Replicates at the bench.
What are replicates?
The term replicate refers to the act of measuring a single variable multiple times to account for variation. There are two types of replicates that are of most importance to preclinical scientists: biological and technical replicates. See definitions below.
- Biological replicates: Taking measurements from several experimental units assigned to an intervention. Biological replicates consider biological variability between experimental units. An example of an experimental unit in an animal study would be the animal itself.
- Technical replicates: Taking multiple measurements on the same experimental unit (e.g. taking multiple sections of brain tissue from a single mouse that was treated with an experimental therapy). Technical replicates tell you how accurate your measurement is within the single experimental unit. In the example, our experimental unit is the single mouse and the multiple sections of brain tissue are technical replicates. Another example would be a measurement in a 96-well plate where plasma from one mouse loaded into three wells. In this case the experimental unit is again the mouse while the technical replicate are the three wells that have been loaded.
Why include replicates in your experiment?
Considering replicates within an experiment is one way of addressing variation. For example, biological replicates capture random biological variation among the experimental units. On the other hand, technical replicates capture variation as a result of equipment, user or within the study protocol itself. Both improve the efficiency of statistical testing.
Who should measure replicates?
All researchers, both preclinical or clinical, should consider the benefits of performing technical and biological replicates in their experiment.
When should replicates be taken?
Replicates should be planned prior to initiating the experiment and included in design of outcome analysis. Biological and technical replicates should be distinguished at these steps. Equally as important is clear and transparent reporting of biological versus technical replicates in manuscripts
Optimal Practices
You are interested in the therapeutic effects of an oncolytic virus on tumor infiltrating lymphocytes. Below are two examples where technical and biological replicates are considered.
Example A)
10 mice with tumors are randomized to receive one of two anti-cancer drugs (oncolytic viruses). One week after treatment, each tumor (1 per mouse) is collected and cut into 12 sections. 3 images are taken per section.
Example B)
10 mice with tumors are randomized to receive one of two oncolytic viruses. One week after treatment, each tumor (1 per mouse) is collected and cut into 6 sections, each section is then placed into individual wells for analysis.
Explanation A)
There are n=10 experimental units (mice) and n=3 x 12 = 36 technical replicates (# of images x # of tumor sections). This can be reported in the manuscript as, “biological replicates were the individual mice in all experiments. Data from each mouse were used in all analyses, n = 10 mice per group. Data for each mouse is the average number of cells per image from each microscopy section (3 images per tumor section X 12 sections)”.
Explanation B)
There are n=10 experimental units (mice) and n=6 technical replicates (tumor sections). This can be reported in the manuscript as, “biological replicates were the individual mice in all experiments. Data for each mouse is the average number of lymphocyte cells per tumor section averaged across all 6 tumor sections.”
Suboptimal Practices
Example A)
Replicates are not considered throughout the duration of the experiment. The tumour from one mouse is cut into 6 pieces and only one piece is placed in an individual well to be studied.
Replicates are considered, but not reported clearly in the manuscript. The manuscript states “Each point on the graph represents the average number of cells per tumor per mouse, n = 10 mice per group.” The reader therefore does not know that one single data point is actually the average of 36 data points and is highly representative of the tumour and has a low level of bias.
Suboptimal practices can arise in two ways:
A) by not considering replicates within the experiment
B) by not clearly reporting replicates in the manuscript.