Commit 2998b866 authored by Jerome Waldispuhl's avatar Jerome Waldispuhl
Browse files

tmp ref 1

parent 3a90497c
......@@ -106,34 +106,31 @@ One of the main conclusions (Discussion, p14) is that intermediate GC content al
\begin{response}{
The effect of GC is seen in Fig 4a, for example, where the yellow curve for GC = 0.5 is the highest. But this could have been shown much more simply by just generating many random sequences with a given GC content and measuring what fraction of these sequences has a multi-branched structure. I am not clear why the number of mutations k is important for this point.}
\experimentstag We already performed the experiment suggested by the reviewer in Figure~6 (dotted lines). There, we uniformly sampled sequences and separately plotted the MFE of structures with and without multi-loops (first row), and frequency of multi-branched structures (second row). We note that the average MFEs are comparable and high (i.e. less stable). Moreover, multi-branched structures also appear to not be uncommon in uniform samples.\\
However, it was not clear that the same phenomenon holds true if we focus our analysis to low energy structures (i.e. the ones that are stable enough to carry functions). For this reason, we used \rnamutants to show that sampling shows that the ensemble of low energy structures is enriched with multi-branched structures at GC content bias of 0.3-0.5. Interestingly, it is not the case in the immediate vicinity of random sequences. Instead, mutants have to significantly move away from the initial seed to increase the likelihood to discover stable multi-branched structures.\\
However, it was not clear if the same phenomenon holds true when we focus our analysis to low energy structures (i.e. the ones that are stable enough to carry functions). For this reason, we used \rnamutants to show that sampling shows that the ensemble of low energy structures is enriched with multi-branched structures at GC content bias of 0.3-0.5. Interestingly, it is not the case in the immediate vicinity of random sequences. Instead, mutants have to significantly move away from the initial seed to increase the likelihood to discover stable multi-branched structures.\\
Variations of the number of mutations k is used to study properties of the local and global neighbourhoods.
\end{response}
\begin{response}{
In general, I am confused by the concentric ring structure for the sampling. I agree that if we take one starting sequence w0, then the properties of the sequences at distance k from this point will depend on w0. However, if we average over many starting points, it is not clear why the distance k should matter. Every sequence will contribute to every ring after averaging over the starting points. So why are the results in 3a 3b 4a etc not independent of k? The answer to this question has something to do with the fact there is a different normalizing Z for each ring and for each starting point. I am not sure of the validity of normalizing these rings separately for every starting point. I cannot understand this either from the statistical physics viewpoint (this is not really a proper thermodynamic ensemble) or from the biological viewpoint (there is not a true model of mutation and selection). This needs to be justified or motivated better, because many of the results depend on the way this is done.}
\hypothesistag Variations of the number of mutations k is used to study properties of the local and global neighbourhoods.
We aim to compare the properties of the local neighbourhoods of random sequences vs global distribution. Although, we acknowledge that it has not been explicitly stated. Instead, we assumed it could be deduced from our description page~???.
\experimentstag Our approach is described on page ???, and the effect of varying sizes of the neighbourhoods is investigated in Figure~6. In short, we ...
\hypothesistag Variations of the number of mutations k is used to study properties of the local and global neighbourhoods. The impact of varying sizes of the neighbourhoods is discussed in Section~2.4 and eventually illustrated in Figure~5. At large mutational distances, the occurrence of stable multi-loops could be attributed to larger diversity of available RNA architectures (defined as shape coverage in \cite{??}). Figure~6 (second row) does not support this claim in the uniform model (dotted lines). By contrast, the frequency of stable multi-loops increases with the growth of structural diversity (plain lines). Although, we are puzzled by this comment because we do not normalize rings as suggested by the reviewer. \\
We hypothesize that this misunderstanding may originate from the description of \rnamutants in Section~2.1 and Section~4.2, which has been used to identify stable structures available in the energy landscape. Eventually, this misunderstanding appears to be related to typos in Section~4.2 that were kindly reported by the reviewer (see below), and most likely prevented a complete understanding of our techniques. We believe that improving the clarity of Section~2.1 and fixing typos in Section~4.2 will address these concerns.
\end{response}
\begin{response}{
It is said that selection for low folding energy drives sequences towards high GC content and low frequency of multi-branched loops. So what if you select for high folding energy (i.e. less negative)? Is this sufficient on its own to generate large numbers of multi-branched loops?
I would suggest the following simple sampling method would be useful:
- generate many sequences at random with GC content in each of the five bins from 0.1 to 0.9.
- find the MFE for each sequence and then bin them according to MFE using 5 bins.
It is said that selection for low folding energy drives sequences towards high GC content and low frequency of multi-branched loops. So what if you select for high folding energy (i.e. less negative)? Is this sufficient on its own to generate large numbers of multi-branched loops?\\
I would suggest the following simple sampling method would be useful:\\
- generate many sequences at random with GC content in each of the five bins from 0.1 to 0.9.\\
- find the MFE for each sequence and then bin them according to MFE using 5 bins.\\
- Now measure properties as a function of MFE and GC content and plot frequencies of multibranched loops on this 5 x 5 array of bins. Does this depend on both properties or just on MFE?}
We believe that Figure 6 already address this question.
\claimstag We do not apply, neither suggest, that any selection mechanism is at play. On the contrary, we showed with \maternal that it would result in lower structural diversity and most importantly prevent the discovery of multi-branched structures. We also believe that experiences on random sequences (uniformly sampled) in Figure~6 already address these comments.
\end{response}
\begin{response}{
Section 4.1.1 - The notation S\_t+1 appears at the end of this section without definition. Should this be P\_t+1?}
Yet is has been fixed in the manuscript.
The reviewer is correct. It is has been fixed in the manuscript.
\end{response}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment