|Thread||Thread Starter||Forum||Replies||Last Post|
|SOM (Self-organising Maps) to identify expression trends in time-course data||rpauly||Bioinformatics||11||02-08-2015 04:00 PM|
|Gene expression and lipids, time-series analysis||sindrle||Bioinformatics||5||08-02-2014 07:24 AM|
|EdgeR GLM linear regression of time series||bsuac6||Bioinformatics||3||03-28-2013 02:50 PM|
|03-16-2017, 04:42 AM||#1|
Join Date: Feb 2017
Periodic versus Linear Trends in Gene Expression Time Series Plots?
Could cyclically reoccurring periodic osculation patterns and features - no matter how transient - carry more informational value when inferring functional and regulatory aspects based on time series plots compared to potentially underlying linear trends to which most of us have limited most of our attention in the past?
When looking up the same pathway for the attached 81 Time Points (TP) and 8 lifespan datasets it can be clearly shown that the 81 time points for measuring the transcription 26 times per hour. This is the only way to ensure that the time span between subsequent transcriptome measurements would never exceed 3 minutes. Only then even very brief - but nevertheless highly distinctively regulatory and functionally relevant osculating periods could be considered, especially when otherwise too many plots would look the same.
As the gaps between time points keep rising the correlations between time series curves belonging to the same GO terms keeps gradually declining until plots of the same GO term or pathway no longer appear more correlated and similar to one another than to the remaining genome. Exactly, when this point is reached no more functional and regulatory inferences should be based on time series plots. Will therefore inferences not considering relevant cyclically recurring oscillation periods always be wrong, especially if this would cause too many plots to look too much the same?
The data, from which the attached microarray time series plots have been drawn using R, comes from:
It has the title:
Dynamics of two oscillation phenotypes in S. cerevisiae reveal a network of genome-wide transcriptional and cell cycle oscillators
The publication about this dataset is:
Chin SL, Marcus IM, Klevecz RR, Li CM. Dynamics of oscillatory phenotypes in Saccharomyces cerevisiae reveal a network of genome-wide transcriptional oscillators. FEBS J 2012 Mar;279(6):1119-30. PMID: 22289124.
The link to it is:
Therefore, one could hypothesize that regulatory and functional inferences based on time series plots should only be considered if the time between subsequent measurements is much shorter than the briefest - but nevertheless functionally and regulatory relevant - difference in osculation patterns.
The maximally acceptable time span between subsequent measurements, which would still allow to make meaningful functional and regulatory inferences based on time series plot similarities, still needs to be experimentally determined because it depends on the overall duration of the cell cycle for each species.
This, in turn, implies that the cyclical nature of periodically reoccurring oscillation patterns tends to over-shadow and thus disguises any potentially underlying linear gene expression trend over time, to which most anti-aging investigators limited most of their attention.
Another informational dimension to highly interdependent time series plots as described above could be added by dividing the GEDI tool into different expression trend regions for better replicating and validating Janssen's conclusion that the proteome will be less similar to the transcriptome later in life when compared to its beginning.
Gabriel S. Eichler, Sui Huang, Donald E. Ingber; Gene Expression Dynamics Inspector (GEDI): for integrative analysis of expression profiles. Bioinformatics 2003; 19 (17): 2321-2322. doi: 10.1093/bioinformatics/btg307
But when I tried to validate this claim by Janssen I found about as many converging as diverging genes. This requires to carefully reproduce Janssen's understandings and concepts of the terms "divergence" and "convergence" and how they were quantified.
Janssens, G. E., Meinema, A. C., González, J., Wolters, J. C., Schmidt, A., Guryev, V., Bischoff, R., Wit, E. C., Veenhoff, L. M., and Heinemann, M. (2015). Protein biogenesis machinery is a driver of replicative aging in yeast. eLife, 4:e08527+.
Could all this lead to the more general conclusion that functional and regulatory inferences based on time series similarities must be flawed when the measuring time points are to far apart for capturing functionally and regulatory relevant periodically reoccurring transient osculation patterns without which it would have been impossible to discern between otherwise identical plots because this would result in functionally unrelated genes to be erroneously placed into the same group?
Since my text to speech software, on which I am depending because I am almost blind, cannot read out this site to me properly, could you please reply directly to my email, which is [email protected]
Last edited by tfh4; 03-16-2017 at 04:47 AM. Reason: forgot attachment
|aging, bioconductor, functional genomics, time series, transcriptional analysis|