I have RNA-Seq samples from an experiment in which I infected epithelial cells with bacteria. The experiment was repeated identically 3 times and then once more with slightly altered parameters (higher bacterial inoculum and longer incubation time). Each experiment also had certain treatment groups. My goal is to determine DEGs between those groups.
When comparing the gene expression of all samples within their groups (intragroup correlation), some of them are clearly different from the rest of the group. Sometimes they also form two distinct clusters (see examples attached, blue color in the scale bar indicates low difference/high correlation). However, these "outlier" samples do NOT necessarily stem from the experiment with altered parameters.
How should I proceed with my data analysis?
My concern is that I will miss otherwise significantly regulated genes because the samples within each group are so heterogenous.
Yes, you can exclude RNA-Seq samples based on low intragroup correlation, but it is important to consider the underlying causes of the low correlation and the potential impact on downstream analysis.
Intragroup correlation is a measure of the similarity between samples within a group, and it can be used to assess the quality of RNA-Seq data. Low intragroup correlation can be caused by technical issues such as batch effects, library preparation artifacts, or sequencing errors, as well as biological factors such as genetic variation, sample heterogeneity, or differences in tissue composition.
If the low intragroup correlation is caused by technical issues, excluding the affected samples can improve the quality and reliability of downstream analysis. However, if the low correlation is caused by biological factors, excluding samples may introduce bias and affect the interpretation of the results. For example, if the low correlation is caused by genetic variation, excluding samples may result in a biased representation of the population or disease.
Therefore, before excluding RNA-Seq samples based on low intragroup correlation, it is important to investigate the underlying causes and consider the potential impact on downstream analysis. This can involve visualizing the data using multidimensional scaling (MDS) or principal component analysis (PCA) to identify batch effects or other sources of variation, as well as performing differential gene expression analysis to assess the impact on the results.
Overall, while low intragroup correlation can be an indicator of poor quality RNA-Seq data, it is important to carefully evaluate the causes and potential impact on downstream analysis before excluding samples.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey