Let say I have 3 treatments and each treatment has 5 replicates, do I prepare 15 cDNA libraries, with 5 individual libraries for each treatment? Or just pool the replicates within each treatment, e.g. 5 pooled samples in each treatment, resulting in 3 cDNA libraries representing 3 treatments?
Does choosing the latter (pooling replicates into 1 cDNA library) makes the results unusable? Or does it provide a better representation of the treatment? Will individual variation in the first method (of preparing individual cDNA library and sequencing for each individual) be an added benefit for analyzing differential gene expression between treatments?
Also, what if funding is a limiting factor? should I just resort to sequencing 2 individuals or 1 individual per treatment instead of pooling?
In my opinion, pooling is NOT a good way to go forward. In that case, you will be missing information on biological variances, which is important for your hypothesis testing. As you mentioned, variance data will be indeed useful for DE gene identification and other downstream analyses.
To save money, I would say, for most experiments 3 biological replicates are good enough though it depends on your particular experimen.
Totally agree with Proyash Roy . Pooling will decrease the quality of your results. What is the point of replicates then? Replicates are done to ensure good quality and to record possible variations. I also agree that 3 replicates would be enough.
Keep in mind that the quality of your work is not about numbers. It is rather about the basic idea and the significance of your study question.
If your question is good enough, you may be able to publish even a single cell transcriptome.
Thanks, Proyash Roy and Iman Hassan Ibrahim for the clarification. I, personally do agree on the individual representation of each treatment and use individuals as biological replicates. However, I have seen several transcriptomic papers with libraries constructed from pooled replicates. Their justification was that pooled replicates would reduce individual variation and allow their data to represent more individuals.
One could argue that we do not really need the data to represent all individuals in one sample.
If we are talking about clinical cases for example, if some genes are DE in 1 patient out of the study cohort, is it wise to consider it DEG? You would find yourself asking: how many patients do you have? That means that the individual variation matters. It is one thing you want to estimate and evaluate, not to conceal.
In your case, the ideal results of the replicates should be the same or almost the same, as your starting materials and all conditions were the same.
Hence, variations could point out methodological errors, so merging would decrease the quality.
My point is : if 1 of your 5 replicates differ than the other 4 in 50% of the expression profile, would you consider this normal? Would it be good if you pool it with other samples?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size