owned this note
owned this note
Published
Linked with GitHub
# Response to EiC Nature Comp. Sci. and Reviewer 1
## Reviewer's comments on revision
> It still seems like the paper falls in the limbo between a computational biophysics and scientific computing audience.
It's interesting that the reviewer uses the word "limbo" because that is where one often feels as a computational scientist! Our work is at a crossing between fields, which is typical in computational science. The paper targets readers who can be computational scientists, computational biophysicists, or computational chemists. It is written in an accessible way for the non-biophysicist computational scientist to follow the model-building and software design, and appreciate the resulting combination of high-performance computation and a high-productivity interactive problem-solving environment. It presents thorough evidence of convergence with respect to the spatial discretization, and correctness of the implementation, to satisfy the readers from the computational biophysics area. We added in revision new results (using third-party software, trusted in these communities) to satisfy the referees that our implementation is correct and equivalent to those other tools. Admittedly these new results do target a subset of readers who are active in computational biophysics (others might be satisfied with the original set of results). But beyond the domain of application, the main message of the paper is with respect to the power of combining high performance (computing) and high productivity (researcher). This message is of general interest to computational scientists.
We belive the above is communicated well in the paper. If the editor suggests that we make it even more clear somehow, we will be glad to try.
Apart from this, the assessment of Reviewer 1 is that "this paper is publishable as-is."
### Additional comment from the editor
> [Reviewer 1] suggested that another example calculation could be provided for the computational biophysics readers, such as the change in the binding with respect to salt or pH (linkage analysis), the influence of mutation on binding, or examination of titration state changes. Is this something you would be willing to do? Would this be feasible? This reviewer also mentioned that there are some PB benchmarks for ligand binding energies, small molecule
solvation, etc. that could be used instead, if this makes more sense.
We heartily disagree that any of this is relevant. A paper cannot be all things to all people. We focus on one application of the PB model, and show that the implementation in software that we provide is _correct_ and equivalent to other third-party software in terms of results. Our workflow, in contrast with other tools, is immediately usable from Python and a Jupyter notebook, making researchers more productive in this application, while at the same time providing high-performance computation.
The suggestion here is almost asking us to write another paper: set up a different problem context, make new calculations, and obtain evidence of correctness in this new setting, presumably comparing with trusted community software (again). This is an unreasonable request and truly outside the scope of our work. Moreover, the PB model has already been successfully used in those contexts by other researchers, and redoing those experiments does not really contribute to the point made in this paper.