GWIPS-viz Forums

Full Version: RUST for dwell time
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Hi,
I am doing my first steps in RP analysis and I find your galaxy platform to be a very useful solution which allows a relatively smooth entry into a complex field. Thank you.
I am trying to use RUST to assess “dwell time” per codon in a sample of interest.
If I understand correctly, I should be using the metafootprint analysis feature at the codon level.
The output I get are two plots as in figure 5 in the nature communication paper (https://www.nature.com/articles/ncomms12915).
The first plot shows RUST ratio (log2) and is useful to decide on read length and offset for A site assignment.
The actual data for this plot is available as another output file.

I believe the second plot is what I need, in that the values per codon should represent dwell time. My questions are:
1. The Y axis for the second plot states: "RUST ration relative to min". what does relative to min means?
2. Where can I find the actual values used for this plot. I was expecting to get a table like for I got for the first plot.
3. Alternatively, can I use the values from the data table of the first plot to calculate the values for the second plot
Many thanks,
Amir
Hi Amire,

The "RUST ratio relative to min" are just RUST ratio of the A-site normalised by the lowest RUST ratio.
The RUST ratios typically vary from 0.2 to 5. The normalisation the RUST ratio of each codon by the lowest RUST ratio obtained from that dataset ensure that the lowest value obtained is fixed at 1.0. This allows for easier comparison between datasets and to gauge the fold range in the differences in codon dwell time.

You are correct in that you can obtain the values for the plot from the data table.
The csv file contains the RUST ratios for many positions across the ribosome, including the ribosome A-site.
The first column (A) is the codon
The 2nd column (B) is the expected RUST value.
The other columns is the observed RUST value at different positions to the A-site.

You should divided the observed RUST value by the expected RUST value to calculate the expected enrichment.
The observed RUST values for the A-site is recorded at the 43rd column (AQ).



Regards
Patrick
Thank you Patrick for the quick response.
I kind of figured it a few minutes before i got your answer.
I have a few more questions which open a broader discussion. I list them here but if you prefer, can post them elsewhere.

1. To what extent do you think this value indeed relates to dwell time. meaning what is it ignoring or assuming.
2. is there a different normalization or a different strategy you would consider in order to look for change in dwell time between samples rather than comparing codons within one sample?
3. it seems that the governing parameter in this plot is the amino acid and not the codon. As if the limiting step is the supply of amino acids. is this indeed the case?
Thanks,
Amir
(20-Jul-2017 01:09 PM)Amire Wrote: [ -> ]1. To what extent do you think this value indeed relates to dwell time. meaning what is it ignoring or assuming.

Most of normalization approaches would give a reasonable estimation of relative dwell times if data were ideal as evident from simulations. RUST advantage is its resistance to noise in comparison with other methods. It seems that there is a lot of noise in the ribo-seq data for various reasons (both biological and technical), so foremost RUST is intended to be practical. The transformation that provides robustness to RUST, however, results in the loss of some useful information. One problem that is associated with that is that RUST consistently overestimates the dwell times of the quickest codons. In other words, the smallest relative dwell times are likely to be even smaller than it is estimated with RUST.

(20-Jul-2017 01:09 PM)Amire Wrote: [ -> ]2. is there a different normalization or a different strategy you would consider in order to look for change in dwell time between samples rather than comparing codons within one sample?

We haven't perform a comparison of approaches for this purpose, it seems that this is simple to achieve as evident from numerous publications where synthesis of specific amino acids was impaired. Perhaps RUST advantage would be less obvious when you compare two datasets obtained from the same biological sample and under the same protocol, as the noise would be similar. But I suspect that RUST could be used for this purpose as well.

(20-Jul-2017 01:09 PM)Amire Wrote: [ -> ]3. it seems that the governing parameter in this plot is the amino acid and not the codon. As if the limiting step is the supply of amino acids. is this indeed the case?

The data are plotted for amino acids on axis x and this is done primarily for convenience. In some cases the variation in dwell times of synonymous codons for a specific amino acids seems substantial. But we do not know how well these data reflect the reality, for example because CHX have a preference for blocking certain codons, see Figure 5 in Lareau et al paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4052883/. It is possible that CHX preferentially acts on amino acids rather than codons, perhaps this was even addressed already and the answer could be found in the literature.
Reference URL's