From:Dianne Cook Sent: Wednesday, April 1, 2020 5:32 PM To:Hsieh, Nan-Hung Subject: Decision on RJ 2019-153
Dear Nan-Hun,
We regret to inform you that you article "pksensi: A Package to Apply Global Sensitivity Analysis in Physiologically Based Kinetic Modeling" submitted to the R Journal has been rejected. The editorial board has discussed your submission and based on the reviews (below) have decided that it is not a suitable contribution for the journal.
We wish you all the best and hope you find the reviews useful.
Regards,
Di
REVIEW 1
Overview: Is the article important to R the community? Is the approach sensible? Are technology and methods up-to-date?
In terms of methodology, the authors added 'random phase shift' to assess sensitivity index (Sobol index) in GSA, they pretended that they could solve the problem of reliability (robustness) due to a small size of samples. This combined with eFAST (a known algorithm), they mounted a package for the assessment of sensitivity of outputs to model parameters. This package will be useful to the R community.
Article: Is anything unclear? Is anything missing? Are the examples helpful? Is there sufficient background, including alternative approaches in R?
I have some concerns about PK examples:
1. In fact, Example 1 is not a PBPK model. It should be simply called a compartment PK model as that used in literature. There are fewer physiological components expressed in this model.
Moreover, the total volume distribution should be written as Vdist/F, where F is the so called bioavalability, but here the author can assume that F=100% for the sake of simplicity. So I suggest the authors to use the general PK terminology to avoid issues raised for the definition of PBPK,
where many organs and tissues should be modeled and integrated into a whole body model. And 'intake' is not a standard term, it should be replaced by 'administration', etc.
In this example, they can use central compartment and depot compartment as that used in NONMEM. Then the code will be simpler.
2. The authors can separate the technical details and the examples. Otherwise it is too redundant to read as they put all together, from R commands, models, results and analysis of the results (pharmacology as well as R codes).
3. It would be much clearer if the authors can provide the illustration of workflow.
-------------
REVIEW 2
Overview
Global SA has a steady stream of attention for a long time, and many packages are developed in R. �sensitivity� is a widely used R package. �pksensi� is developed for assessing reliability, and provides the plot for checking convergence as well as sensitivity index.
1. The main point of this paper is that the authors modified �fast99� function in "sensitivity" library to �rfast99� with replication. This replication is for calculating the convergence index. SI index and CI index are calculated for global SA. To represent the result of these indices, �pksensi� provides �heat_check,� �pksim,� and �plot.rfast99� function for plot and �check� for the summary result.
Even though this modification for replication and heat_map for representation are useful, it is possible to be done with fast99 function in "sensitivity" library with a little effort to replicate. Also, the user cannot use many other methods in �sensitivy� package, which are not provided in �pksensi.�
2. �rfast99� class definition is not clear, and the class of the result from �solve_mcsim� function is not clear. In �https://github.com/nanhung/RJ-pksensi/blob/master/RJ-pksensi.R,� the class of the result in line 122 is �rfast99� ( in example 1). However, the class of the result in line 180 (in example 2) is �array,� not �rfast99.� The result from the same function should have the same class.
1. Even though the calculation of SI is from "sensitivity" library, SI is an important index in this paper, and it is better to write how to calculate SI briefly in this paper with the reference.
2. The definition of CI is not clear. In equation on the bottom on page2, what is the meaning of "max"? Max of replications? It is not clear.
3. Typo on the first line of page3. Second $SI{I,t}^{ub}$ should be $SI{I,t}^{lb}$
4. In Equations in example 1, there is no equation for "dAmetabolized."
5. On page 5, the name of the parameter is confused. "ka" means "kgutabs"? "Kelim" means "Ke"? It should be consistent.
6. There is no function for Figure 6.
7. The output of the function "check" is not clear. In "sensitivity check" and "convergence check", what are the meaning of "First order:", "Interaction:", "Total order:", and "unselected factors in total order:"?
8. In example 2, the output of "solve_sim" function is not the class of "rfast99". Therefore, plot, print, and check functions do not work.
------------------------
Package
1. The output from rfast99, solve_fun, and solve_mcsin is the object with 'rfast99' class. However, the structure of output from rfast99 is different from the others.
2. The author needs to explain the structure of "rfast99" class in detail.
3. The help files for "check," "heat_check," "plot," and "print" should be separated and explained in detail.
4. The structure of rfast99 in "pksensi" is very similar to the structure of fast99 in "sensitivity." Is there any specific reason to define a new class "rfast99" instead of using "fast99"?
%>%>%>%>%
Professor Dianne Cook
Executive Editor, R Journal dicook.rj@gmail.com
Sent by Hsieh, Nan-Hung (nhsieh@cvm.tamu.edu). Created by fire.
Sent from Mail for Windows 10
From:Hsieh, Nan-Hung
Sent: Wednesday, April 1, 2020 5:36 PM
To:nanhongg.698fba@m.evernote.com
Subject: FW: Decision on RJ 2019-153
Sent from Mail for Windows 10
From:Dianne Cook
Sent: Wednesday, April 1, 2020 5:32 PM
To:Hsieh, Nan-Hung
Subject: Decision on RJ 2019-153
Dear Nan-Hun,
We regret to inform you that you article "pksensi: A Package to Apply Global Sensitivity Analysis in Physiologically Based Kinetic Modeling" submitted to the R Journal has been rejected. The editorial board has discussed your submission and based on the reviews (below) have decided that it is not a suitable contribution for the journal.
We wish you all the best and hope you find the reviews useful.
Regards,
Di
REVIEW 1
In terms of methodology, the authors added 'random phase shift' to assess sensitivity index (Sobol index) in GSA, they pretended that they could solve the problem of reliability (robustness) due to a small size of samples. This combined with eFAST (a known algorithm), they mounted a package for the assessment of sensitivity of outputs to model parameters. This package will be useful to the R community.
I have some concerns about PK examples:
1. In fact, Example 1 is not a PBPK model. It should be simply called a compartment PK model as that used in literature. There are fewer physiological components expressed in this model.
Moreover, the total volume distribution should be written as Vdist/F, where F is the so called bioavalability, but here the author can assume that F=100% for the sake of simplicity. So I suggest the authors to use the general PK terminology to avoid issues raised for the definition of PBPK,
where many organs and tissues should be modeled and integrated into a whole body model. And 'intake' is not a standard term, it should be replaced by 'administration', etc.
In this example, they can use central compartment and depot compartment as that used in NONMEM. Then the code will be simpler.
2. The authors can separate the technical details and the examples. Otherwise it is too redundant to read as they put all together, from R commands, models, results and analysis of the results (pharmacology as well as R codes).
3. It would be much clearer if the authors can provide the illustration of workflow.
-------------
REVIEW 2
Overview
1. The main point of this paper is that the authors modified �fast99� function in "sensitivity" library to �rfast99� with replication. This replication is for calculating the convergence index. SI index and CI index are calculated for global SA. To represent the result of these indices, �pksensi� provides �heat_check,� �pksim,� and �plot.rfast99� function for plot and �check� for the summary result.
Even though this modification for replication and heat_map for representation are useful, it is possible to be done with fast99 function in "sensitivity" library with a little effort to replicate. Also, the user cannot use many other methods in �sensitivy� package, which are not provided in �pksensi.�
2. �rfast99� class definition is not clear, and the class of the result from �solve_mcsim� function is not clear. In �https://github.com/nanhung/RJ-pksensi/blob/master/RJ-pksensi.R,� the class of the result in line 122 is �rfast99� ( in example 1). However, the class of the result in line 180 (in example 2) is �array,� not �rfast99.� The result from the same function should have the same class.
The result of line 122
+ vars = outputs, time = t,
+ condition = conditions))
Starting time: . . .
Execute: ./mcsim.pbtk1cpt.model.exe sim.in
MCSim v6.1.0
.
.
.
[1] "rfast99"
The result of line 180
+ monte_carlo = 1000, dist = dist, q.arg = q.arg,
+ time = times, condition = conditions,
+ rtol = 1e-7, atol = 1e-9)
Starting time: . . .
Execute: ./mcsim.pbpk_apap.model.exe sim.in
MCSim v6.1.0
.
.
.
[1] "array"
--------------------------------------------------
Article
1. Even though the calculation of SI is from "sensitivity" library, SI is an important index in this paper, and it is better to write how to calculate SI briefly in this paper with the reference.
2. The definition of CI is not clear. In equation on the bottom on page2, what is the meaning of "max"? Max of replications? It is not clear.
3. Typo on the first line of page3. Second $SI{I,t}^{ub}$ should be $SI{I,t}^{lb}$
4. In Equations in example 1, there is no equation for "dAmetabolized."
5. On page 5, the name of the parameter is confused. "ka" means "kgutabs"? "Kelim" means "Ke"? It should be consistent.
6. There is no function for Figure 6.
7. The output of the function "check" is not clear. In "sensitivity check" and "convergence check", what are the meaning of "First order:", "Interaction:", "Total order:", and "unselected factors in total order:"?
8. In example 2, the output of "solve_sim" function is not the class of "rfast99". Therefore, plot, print, and check functions do not work.
------------------------
Package
1. The output from rfast99, solve_fun, and solve_mcsin is the object with 'rfast99' class. However, the structure of output from rfast99 is different from the others.
2. The author needs to explain the structure of "rfast99" class in detail.
3. The help files for "check," "heat_check," "plot," and "print" should be separated and explained in detail.
4. The structure of rfast99 in "pksensi" is very similar to the structure of fast99 in "sensitivity." Is there any specific reason to define a new class "rfast99" instead of using "fast99"?
%>%>%>%>%
Professor Dianne Cook
Executive Editor, R Journal
dicook.rj@gmail.com