This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Non-linear relationship between a cell's mitochondrial content and detection of ribosomal genes ssen in single cell rNA seq

Hi, could anyone explain why there is an inverse relation between the percentage of mitochondrial (-mt) genes of a cell and its expression of ribosomal(Rp[s, l]) genes? I have been doing analysis of single cell RNA seq data and I noticed that in general that cells with a high ratio or percentage of Rp[s, l] genes tend to have a low ratio/pecentage of -mt genes.

Could there be any biological explanation for this? See code and figure below:

ggplot(data = metadata, aes(x = mitoRatio, y = riboRatio)) + 
  geom_point(color='blue') +
  ylab("ratio of Rp[s, l] genes")+
  xlab("ratio of -mt genes")+
  geom_smooth(method = "lm", se = FALSE)+
  theme(aspect.ratio = 1)

ratio of ribosomal protein genes in a cell vs ratio of mitochdrial genes per cell

rna-seq next-gen
• 2,383 views • link
updated by Ventrilocus Ventrilocus • written by charlesgwellem • charlesgwellem

Can you elaborate:

  • what do you classify as ribosomal gene? Is it the type being rRNA (e.g. the type column in a GTF) or ribonucleoproteins (Rp[s,l])?
  • with mithcondrial content you mean the % of reads aligned to genes that probably are "^mt"?
  • can you show data, like a plot to make the issue clear, I personally do not like commenting on textual descriptions as they are highly implicative and I prefer to interpret data myself before making a comment.
• 0 views • link
written by ATpoint ATpoint

It is still unclear what exactly mitoRatio and riboRatio. Is this correlation significant (cor.test)?

Edit: It is the full link you have to paste into the image putton field, I did that now, please stop editing.

• 0 views • link
written by ATpoint ATpoint

Thanks for the correction. The riboRatio and mitoRatio correspond to the percentages of reads that map to ribosomal protein genes and mitochondrial genes respectively. But I chose to represent them as ratios.

The equation of the line is y = 0.0588 - 0.0603x, r^2=0.168 and the pvalue from the cortest result shows significance.

> cor.test(metadata$riboRatio, metadata$mitoRatio)

    Pearson's product-moment correlation

data:  metadata$riboRatio and metadata$mitoRatio
t = -14.713, df = 1071, p-value < 2.2e-16
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
 -0.4586427 -0.3590156
sample estimates:
       cor 
-0.4100516
• 0 views • link
written by charlesgwellem • charlesgwellem

1 answer

Assuming that:

  1. mitoRatio = # of mitochondrial genes observed/total number of genes observed
  2. riboRatio = # of rRNA genes observed/total number of genes observed (maybe you also included ribosomal proteins to this ratio)

If you go to the right extreme of the x-axis, you reach more than 75 % of mitochondrial genes (mitochondrial genes are almost exclusively observed). There are only 2 rRNA genes out of the 37 mt genes. Since there are only 37 mt genes, it is much more likely to observe most of them in a scRNA experiment. The expected riboRatio is expected to be around 5.4 %. On the left extreme, almost all genes observed are nuclear. Though there are many more nuclear genes compared to ribosomal genes, the very high expression of rRNAs makes them much more more likely to be observed compared to other genes that are either not expressed or residually expressed (hence, achieving on average larger riboRatios than mitochondrial read-enriched cells). The continuous variation of the ratio of mt to nuclear reads could give rise to a spurious negative correlation between mitoRatio and riboRatio. Even if the definitions assumed here are not correct, I am quite certain that the observed relationship derives from the enrichment in mtRNA reads.

• 0 views • link
written by Ventrilocus Ventrilocus

Log in to answer this question.

More posts like this