Estimating mRNA Half-Life from Single Time Point Data Using mRNA/Ribo-seq Expression Data

Question

[Ramya]

Hi all,

I'm currently working on a project where I need to estimate mRNA half-lives using expression data from a single time point. My dataset includes both mRNA expression data and ribosome profiling (Ribo-seq) data. I've read about methods that use regression analysis and other computational approaches to estimate mRNA half-life from single time points, but I'm looking for more specific guidance or tools/scripts that could help with this task.

Specific Questions:

1. Are there any recommended tools or scripts for estimating mRNA half-life from single time point data using mRNA or Ribo-seq expression data?
2. Can mRNA half-lives (t1/2) be estimated in human cells using only mRNA and ribosome profiling (Ribo-Seq) data from a single time point?
3. Can established reference datasets of mRNA half-lives serve as control points for comparative analysis?
4. Could regression-based techniques, utilizing features derived from mRNA and Ribo-Seq data (e.g., transcript abundance, translation rates, codon usage), be employed to predict t1/2?
5. Are there publicly available reference datasets of human mRNA half-lives suitable for such analyses?

Answer 1

[i.sudbery]

Its not possible to estimate mRNA half-life from standard mRNA-seq, unless the RNA has been treated with a metabolic label, or transcription has been inhibited. Its not about lacking time points - its just that mRNA-seq data caputres steady state RNA levels, not dynamics.

I'm also, say 90% sure that you can't calculate RNA half-life from Ribo-seq data either, although I don't know the protocol in detail enough to be 100% sure.

[Jack Tierney]

Ribo-Seq can be seen as just a form of selective RNA-Seq where the fragments of RNA that are sequenced are enriched for those that were within the ribosome at the time of digestion with an RNase. So for the same reasons you mentioned for RNA-seq it is not possible with Ribo-Seq either.

[i.sudbery]

I did just briefly considered if some of the ways in which you treat cells during the riboseq protocol might inhibit transcription.

[Jack Tierney]

I was more trying to 'yes and...' your answer than imply you did not know that for yourself.

My understanding is that inhibiting transcription would induce stress response and as a result confound the conditions you are trying to observe translationally. So something that co-inhibits transcription and translation would not be favourable. Furthermore, I am unclear on how just inhibiting transcription could lead to these insights without, as you said, metabolic labelling. Interesting to think about all the same.

[Ramya]

Thank you so much for taking the time to respond to my question! I understand what you're saying, and I have edited my question to be clearer. If you have any further suggestions or insights, they would be greatly appreciated.

[i.sudbery]

1. Are there any recommended tools or scripts for estimating mRNA half-life from single time point data using mRNA or Ribo-seq expression data?
2. Can mRNA half-lives (t1/2) be estimated in human cells using only mRNA and ribosome profiling (Ribo-Seq) data from a single time point?

I'll take these two together (as they seem to be more or less the same question. There are no tools that can estimate mRNA half-life from single time point data. Its not really clear what you mean here by single time point. If there were multiple time points, they would be time since what? It it was time since transcription inhibition say, then multi-time point data could be used to estimate transcript half-life. But even with transcription inhibition you would need at least two time points to even try an analysis, and most approaches to this in the past have used lots more time points. With a highly dense time course it might even be possible to estimate half-life from the shape of the expression curve where transcription is increased, although i'm not aware of anyone who has ever tried this. But in general RNA-seq and ribo-seq just do not contain any information that is relevant to RNA half-life.

1. Can established reference datasets of mRNA half-lives serve as control points for comparative analysis?

If you had, say a set of reference transcript levels for a cell type, and then you had inhibited transcription for, say an hour, one could see who it would be initially plausible to compare how much transcript levels have decreased in that hour. Unfortunately RNA-seq is far too variable from sample to sample to make this possible, even for differential expression. For estimating a half-life (which you'd normally want 3-6 time points), it would be hopelessly sensitive to difference between experiments.

1. Could regression-based techniques, utilizing features derived from mRNA and Ribo-Seq data (e.g., transcript abundance, translation rates, codon usage), be employed to predict t1/2?
2. Are there publicly available reference datasets of human mRNA half-lives suitable for such analyses?

Here the answer is a little bit more nuanced. There have been attempts to predict mRNA half-life from features such as translation rates, codon usages, motif presence, RBP binding etc. The best examples of this manage to get an R2 of about 0.4 between their predicted half-life and the measured half-life (see for example https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02811-x), meaning that they can explain about 40% of the variance in half-life from sequence. But even this only works when the model is specifically trained on the cell type you are using. The paper I reference in the las sentence has a good compendium of the available reference data sets on mRNA halflife.