Discuss exactly what should be in the paper

[lmoffatt]

My original idea is to fit the same models of Moffatt & Hume plus some models based on the binding site being on the interface.

[lmoffatt]

The first part is straighforward: I set the priors around the values obtained in Moffatt&Hume.

Now to compare to the new models, thats demands some thinking on how to decide the priors.

However, it is easy to try different prior schemes, but not so much different parameterizations.

[lmoffatt]

As now nothing changed in the objectives. We have a system that works for scheme 4 with all the data, although it did not equilibrate ate completely, we need 4 days using 64 cores for that. So, it is necessary to do three things:

1. re-implement the allosteric models
2. implement the continuation
3. optimize the code, the low hanging fruits.

[lmoffatt]

Now I am in a point were I want to write the paper. So, what should be in the paper?

1. Figures
2. Objectives
3. Methods
4. Discussion

What is the objective? What are possible objectives that I had? There are two kinds: related to development of methodology and related to solving actual problems.

The thing is that to really advance in any of them you need the other. Solving problems sometimes require new methodology and to develop new methodology you need to try it on actual problems.

This paper is unusual because I started by the methodology, not, as usual, by the problem. Biologist tend to focus on the problem, probably because focusing in the methodology requires knowledge in other disciplines that are not accessible to the average biologist. On the other hand, people from other disciplines that develop the methodology might not be acquainted with the subtleties of biological thinking. Or many times, new methodologies might have marginal interest. It is difficult to develop a new technology that really moves the needle, probably because there are few of them. So, it requires a lot of effort to understand a new methodology with the perspective of a scarce reward. It is wiser to go for the low hanging fruits.

So, there is a mixture of craziness, unlimited self-confidence and disregard for your academic career in the decision to develop new methodology in Biology.

[lmoffatt]

Possible objectives, unfiltered list: #246

from the methods 1) Do models converge? 2) Do they converge to the same value? 3) If we simulate with the value they converge, do we recover the same value? 4) Is E[FIM]== -Var[Gradient], test using simulations or posterior likelihoods.

to understand the value of Macror: 5) Does Macror provides information not present in MacroNR ? (for instance in pre-jump) 6) what is the scheme with the greatest evidence? 7) Should I try RR and RRR or BRB?

8) which is the posterior probability of the evolution of the states? 9) which is the posterior probability of the kinetic rate constants?

[lmoffatt]

Figures (all ): #247 #248

1) description of schemes

2) logLikelihood vs iterations

3) logP vs iterations

2) plogEvidence vs iterations

4) logEvidence vs iterations

3) {loP, logL, logEvidence) vs beta?

4)

[lmoffatt]

So, schemes 1- 11 up to 50.000 iterations

and continuation 6 for Q0, Qa and Pi and might be Pcov

[lmoffatt]

Everything listed in this issue is either implemented or in new issues