For a standard experiment using RNA-seq you should plan for at least 3, and ideally 4-6, independent biological replicates per group. When working with clinical samples, or when biological variability is high or the expected differences are small, consider increasing the sample size to 8 to 12 samples per group.
A larger number of biological replicates usually improves the reliability of differential gene expression analysis (detection of DEGs) more than sequencing a few samples at very high depth.
How many replicates should you choose?
| Project type | Number of replicates per group | Recommendation |
|---|---|---|
| Pilot experiment | 3-4 | Allows you to estimate variability and effect size |
| Standard RNA-seq | 4-6 | Best choice for most projects |
| Cell lines or material with low variability | 3-5 | At least 3 independent cultures |
| Laboratory-grown plants and inbred lines | 4-6 | Each replicate should be grown independently |
| Animal tissues | 5-8 | Take age, sex and the source of the material into account |
| Human or clinical samples | 8-12 or more | Usually high inter-individual variability |
| Subtle differences in gene expression | 8-12 or more | Statistical power analysis is recommended |
| Time course | 3-6 per time point per group | Each time point requires its own replicates |
Biological vs. technical replicates
Biological replicates are independent organisms, patients, cultures or experiments within the same group.
Technical replicates are generated by repeating the extraction, preparing several libraries or resequencing the same material. They do not replace biological replicates and are usually combined in the analysis as a single sample.
When are 3 replicates sufficient?
Three replicates can be used when:
- the project is a pilot study,
- the material is homogeneous,
- the experimental conditions are well controlled,
- the expected differences in gene expression are large,
- the budget does not allow for more samples.
However, this is the minimum. Losing or excluding one sample leaves only two replicates, which substantially limits statistical analysis.
When should you choose 4 to 6 replicates?
This is the recommended option for most standard RNA-seq experiments, particularly when:
- the aim is differential-expression analysis of genes,
- the expected effects are moderate,
- the results are intended to form the basis of a publication,
- the material shows natural variability,
- you plan to carry out GO, KEGG or GSEA analysis.
When do you need 8-12 samples or more?
It is advisable to plan for more replicates when:
- you are studying clinical samples or samples from different individuals,
- the expected changes in gene expression are small,
- the material is highly heterogeneous,
- the analysis involves multiple factors,
- you are looking for biomarkers,
- pilot data are available for power analysis.
How do you calculate the total number of samples?
Multiply the number of groups, i.e. combinations of genotypes, conditions and time points, by the number of replicates.
Example:
2 genotypes × 2 conditions × 4 replicates = 16 samples
Each combination of conditions should have its own independent biological replicates.
Can samples be pooled?
Yes, if a single organism or tissue does not provide enough RNA. However, bear in mind that:
- one pool counts as one biological replicate,
- individuals combined in a single pool are not separate replicates,
- each pool should be prepared independently,
- all pools should contain comparable quantities of material.
Three independent pools of five plants each represent 3 biological replicates, not 15.
Sample Requirements
All samples should:
- come from clearly defined groups,
- be collected at a comparable stage,
- be prepared and stored in the same way,
- have comparable RNA quality,
- have complete metadata,
- have unambiguous labels.
For standard mRNA-seq, the following is recommended:
- more than 0.5 µg of total RNA,
- a concentration above 50 ng/µl,
- a volume above 20 µl,
- OD260/280: 1.8-2.2,
- RIN above 7, or above 6 for plants.
Low-input samples, FACS-sorted samples or degraded material require case-by-case assessment.
Approximate data output
For standard gene expression analysis, the following are usually recommended:
- 20-30 million read pairs per sample,
- approximately 6-10 Gb of data per sample at 2 × 150 bp.
With a limited budget, it is better to increase the number of biological replicates than to sequence just a few samples at very high depth.
Common mistakes
- using only two samples per group,
- treating technical replicates as biological replicates,
- combining all individuals into a single pool,
- replacing replicates with deeper sequencing,
- not having additional material available in case a sample is excluded,
- preparing controls and experimental samples in different batches,
- failing to account for sex, age, genotype or other sources of variability,
- treating cells from a single donor in single cell and spatial omics as independent replicates.
Plan your project
Unsure and need advice? Describe your project - we will select the appropriate number of samples, method and data output.
Based on your experimental design and material type, we will help you determine the number of biological replicates, the library preparation method and the appropriate sequencing depth.
Selected projects and client publications
See which institutions we perform next-generation sequencing (NGS) for and explore scientific publications that draw on the data we have supplied.