federicogiorgi
commented
3 years ago Dear Eva,
in theory, the principle of co-occurrence will still be valid in time-series data, and so it makes sense to try corto. However, in order to detect time-shifted causality structures (e.g. TF1 upregulated at time point 1, inducing upregulation of target genes observed at time point 3) you could try to provide as input data a combination of two matrices:
- one matrix for TFs (minus the last sample)
- a matrix for target genes (TGs) (minus the first sample) This will pileup up the TGs as if they were measured with a time delay, allowing to detect also TF-->TG relationships that are shifted by one time point. It's the principle behind Granger Causality after all.
Let me know if I wasn't clear :-)
Federico
On Mon, 25 Jan 2021 at 13:00, emaleckova notifications@github.com wrote:
Dear corto-developers,
Thank you for this package and especially the tutorial-like vignettes, which make the first-time use so easy. If I am not mistaken, the example data, such as the MYCN dataset, are patient's data and various treatments. The dataset I would like to analyze with corto comes from a time-course RNA-seq experiment. I do have two treatments and a couple of biological replicates for each time point, but the point of concern is that not all samples from treatment A are "equal" but obviously affect by the time factor.
My question is whether corto can be used with such a dataset at all. If so, can I simply use the normalized expression data or is any additional pre-processing needed? Or would you recommend analyzing only a few selected time points, which I would then identify rather artificially (e.g. at light-dark transition of the 24hr cycle)?
Thank you for your opinion.
Best regards, Eva
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/federicogiorgi/corto/issues/3, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAKWILFLW2KD4Z2N4626JFDS3VMPDANCNFSM4WRUOU5A .
Dear
corto-developers,Thank you for this package and especially the tutorial-like vignettes, which make the first-time use so easy. If I am not mistaken, the example data, such as the MYCN dataset, are patient's data and various treatments. The dataset I would like to analyze with corto comes from a time-course RNA-seq experiment. I do have two treatments and a couple of biological replicates for each time point, but the point of concern is that not all samples from treatment A are "equal" but obviously affect by the time factor.
My question is whether
cortocan be used with such a dataset at all. If so, can I simply use the normalized expression data or is any additional pre-processing needed? Or would you recommend analyzing only a few selected time points, which I would then identify rather artificially (e.g. at light-dark transition of the 24hr cycle)?Thank you for your opinion.
Best regards, Eva