Importance of biological replicates in biological research

 

Tania, a PhD student is conducting experiment on chickpea plants to check a specific gene expression level upon providing phosphorous deficiency stress. She is repeating this experiment for the third time and she could not reach to a conclusion yet, as her data shows specific pattern of gene expression difference among the control and stressed (she got higher expression in the stressed plants) but there is no statistical significance in this difference which is making her data scientifically invalid.

Most of the researcher may have faced similar kind of problem during experiment. This happens mostly when the sample size or sample number is low due to low number of biological replicates, which is also called experimental replicates.

The number of sample sets you are taking during an experiment is called replicates, which indicates for how many samples you are conducting the same experiment? In biological research, these replicates are called 'biological replicates'. For the researcher, the number of sets of the samples matters a lot for a valid statistical analysis of the data obtained from the experiments. A valid statistical analysis proves how much accuracy and robustness are there in your experiment and how reproducible your data is. Reproducibility is an obvious factor for your publication to go in high impact journals.  Therefore, replicates are crucial for scientific research. Popularly, replicates are called N number, which actually reflects the population size or number of individual for statistical analysis. This number is also important to get the statistical significance of the difference of two or multiple data sets or significance of correlation of two data sets. The following points show why high biological replicate numbers are so important for a biological science researchers and how one should chose the replicate numbers.

Replicate numbers ensure the robustness of your experimental data. Suppose, you are replicating your experiments for 10 data sets, and you are getting a specific data pattern for 7 sets and reverse pattern for rest 3 sets, there is higher chance for the data pattern for the first 7 replicates to be significant, even you are including the 3 sets showing reverse pattern. Those 3 sets will be called as outlier which may be resulted due to the experimental error or sample’s impurity.

 

High replicate number helps to reduce chances of repetition of the experiments. If you conduct the experiment once with high replicate numbers then it is likely to get a robust data and there will be no need to repeat the whole experiment for the second time.

 

High replicate number gives you flexibility to remove the outliers. Go back to the first point, where 3 outliers among the 10 data sets were showing reverse pattern. Now you have the flexibility to use the 7 data point providing similar data pattern and to remove three outliers. This will give you higher chance to obtain significant result.

 

There is no fixed number for biological replicates. But for a valid statistical analysis a minimum number of three is required as biological replicates. There is no limit of maximum number for biological replicates.

 

Always try to take more than three biological replicates for the samples which are collected after an uncontrolled treatment. For example, if you are harvesting a plant sample after real insect feeding or you are harvesting field samples for the plants or crops, you cannot control the condition, means you cannot control the hunger of insect or you cannot control the temperature, humidity or soil microbes of the field. In that case you need to take more than three biological replicates, preferably 4 to 10 or more to get robust data.

 

Pooling is a nice practice for collecting biological replicates. You may pool your samples to reduce sample size. For example, if you are doing an expensive experiment ( e.g. RNAseq, or Untargeted LC-MS) where you want to reduce your sample numbers but want to remove batch effect maximally, then you may pool three or more samples as a single sample and collect more two pool similarly. Hence, you can collect (3 samples X 3 pool) = 9 samples but you have to perform three experimental replicates, which will be cost effective for you but chances of hampering data robustness will be less.