6

What's the most widely accepted tool for doing pseudo-temporal ordering from scRNAseq data? Also is there away to separate differential expression that occurs based on "cell identity" or maybe more accurately cell type fate from that which arises from cells being in different stages differentiation.

To be more concrete, lets say there's a population of cells, some of which were born at time 1, time 2, and time 3. The progression along the temporal trajectory can be described via a set of genes that are fluctuating as the cell matures. So you might have the same cell type which was born at time 1 be transcriptionally distinct from a younger one born at time 3. On the other hand, within this population there are subpopulations which will have different cell fates and are transcriptionally distinct. Is there away to reliably separate the temporal axis from the "cell fate axes". If not is this something people are working on or is flawed logic to think this kind of thing is possible?

Alon Gelber
  • 173
  • 1
  • 5
  • 1
    I think this question might be too speculative for this platform. In particular, this involves defining cell types and states, which in many ways is philosophical. – Valentine Svensson May 16 '17 at 21:42
  • What does it mean the tag? it stands for small conditional RNA? I think then that this question has two different issues, one about scRNA and another about separating subpopulations from a pool. Could you clarify it? – llrs May 17 '17 at 07:43
  • 2
    I think the tag stands for single-cell RNA-seq (correct me if wrong?). Although the question is broad, I do know people working on this problem and developing tools - unfortunately I'm not an expert. – Sarah Carl May 17 '17 at 08:02
  • @SarahCarl Maybe you can become one researching an answer for this question :D. If it is too broad, I think you can vote it too, I think it doesn't show research effort, but I am unsure if it is too broad or not – llrs May 17 '17 at 08:30
  • Without being an expert on scRNA-seq it sounds to me like you are looking for something like the monocle package. – mxenoph May 17 '17 at 09:26
  • @Llopis I think it is too broad - but if edited could work well. If this question were instead a question about pros/cons of a few different tools, that might work better. – kevbonham May 17 '17 at 13:12
  • For reference, here is a list with most (but not all) methods available: https://github.com/agitter/single-cell-pseudotime . I don't think this is a question which can have a clear answer without major research. See also this review of strategies: http://onlinelibrary.wiley.com/wol1/doi/10.1002/eji.201646347/abstract . Monocle is probably the most well known, but has been criticized often, leading to people making new tools when it doesn't work. – Valentine Svensson May 17 '17 at 13:25
  • @Valentine thanks for your response, I've seen the github page before but it's good to link it here. And thanks for the review I'll read through it and see what I can glean from it. I wanted some opinions on what method/tool seems like it does the best job at this at this point in time. If anybody has opinions or experiences with this I'd love to hear them. – Alon Gelber May 17 '17 at 16:52
  • Also we'd like to be able to incorporate known expression patterns into the models. – Alon Gelber May 17 '17 at 17:08
  • 1
    @Alon Your question is interesting, but it's sounding like a research project rather than something with a readily available tool :) . If you know your expression patterns are switch-like, have a look at Ouija: https://github.com/kieranrcampbell/ouija. If the patterns are different, try finding a statistician nearby who is familiar with latent variable models. – Valentine Svensson May 18 '17 at 00:32

1 Answers1

3

Regarding the first part of the question: scRNA-seq is a rapidly developing field so it may be hard to talk about "widely accepted tool for doing pseudo-temporal ordering from scRNAseq data". Few of the tools aiming to do just that include Monocole, Waterfall or Sincell (see this paper for references https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4122333/)

The second part is a bit more complex. Many aspects have come together here, e.g. cell-cycles, sub-populaitons identification and pseudo-temporal ordering, to get a true reflection of the biological processes. There are efforts on all (and more) these fronts (see again e.g. the above paper, though not the latest) and there are most likely people working on their integration. I'm not aware of any studies published yet at this depth though

olga
  • 481
  • 2
  • 8