JOSS manuscript comments

[kellypierce]

episimlab/paper.md

Line 26 in ce4b880

These _in silico_ disease models simulate real-world transmission dynamics and are therefore well-suited for such tasks as early detection of novel pandemics, improving situational awareness during periods of high prevalence, and estimating efficacy of intervention strategies.

Change "are therefore well suited to" to "can be applied to" (since the examples listed require additional infrastructure on top of the compartmental model)

episimlab/paper.md

Line 29 in ce4b880

As more data on case incidence, hospitalizations, and viral genomics become available, disease modelers are able to simulate increasingly complex disease dynamics.

Consider changing "viral" to "pathogen" genomics.

episimlab/paper.md

Line 30 in ce4b880

Developing such complex compartmental models is time-consuming, however.

Don't end sentence with conjunction.

episimlab/paper.md

Line 39 in ce4b880

1. Implement any compartment structure that can be represented as a graph

Are there any constraints on the topology of the graph? It can be cyclic, right (e.g. an SIRS model)

episimlab/paper.md

Line 40 in ce4b880

2. Have many input parameters, often with multiple, varying dimensions like age, location, and risk factors

It might be helpful to provide a brief example here, along the lines of "... varying dimensions such as age-depended recovery rates, location-dependent contact patterns, and other demographically stratified risk factors." (but shorten if that seems too long)

episimlab/paper.md

Line 41 in ce4b880

3. Simulate dynamic interventions, such as administering vaccines only when case incidence exceeds a threshold

Change to "Simulate interventions, including dynamic interventions such as ...."

episimlab/paper.md

Line 44 in ce4b880

The package is designed to be approachable; it includes several pre-packaged models that the user can run in a few lines of code.

"approachable" and "easily" are somewhat subjective; maybe replace approachable with "extensible" and omit easily

episimlab/paper.md

Line 45 in ce4b880

When the user chooses to add data streams or more complex transmission dynamics, they can easily do so by adding or replacing processes in the model.

You could reframe as "different or more complex"

episimlab/paper.md

Line 59 in ce4b880

It gives the user flexibility to define key components of their compartmental model, such as the dimensionality of the Markov state matrix, the number of compartments, structure and rate of transitions between compartments, and custom stochastic behavior.

What do you mean by "Markov state matrix" here?

episimlab/paper.md

Line 63 in ce4b880

`ComptModel` implements a Gillespie algorithm that supports Markovian models in discrete time [@gillespie], using a generic model of compartmental disease.

Not sure about this citation (I'll find one from the epi literature to replace). Recommend replacing "Markovian" with "deterministic or stochastic". Also specify that transitions are drawn from a Poisson distribution. (I think the term "Markovian" is correct here, but I don't typically hear these types of models described as such. And I suppose it's technically possible to have a non-Markovian stochastic implementation...)

episimlab/paper.md

Line 70 in ce4b880

Specifically, prototypes of Episimlab were used in studies forecasting hospital burden due to the COVID-19 epidemic in Austin, Texas [@Pierce2020report; @Pierce2020ieee].

Replace "forecasting" with "projecting"

episimlab/paper.md

Line 103 in ce4b880

We would like to thank the UT Austin COVID-19 Modeling Consortium for collaborative support. In particular, we acknowledge Dr. Lauren Ancel Meyers and her team for their guidance throughout development of Episimlab.

Need to add citations for the partition_v1 model and for the vaccine model. (I'll find and add)

[ethho]

Pushed suggested revisions to #52 (9df32d7)

Remaining To-Do

• episimlab/paper.md

  Line 63 in ce4b880

  `ComptModel` implements a Gillespie algorithm that supports Markovian models in discrete time [@gillespie], using a generic model of compartmental disease.

  • Epi citation for Gillespie algorithm/Markovian models

• episimlab/paper.md

  Line 103 in ce4b880

  We would like to thank the UT Austin COVID-19 Modeling Consortium for collaborative support. In particular, we acknowledge Dr. Lauren Ancel Meyers and her team for their guidance throughout development of Episimlab.

  • Mention pre-packaged models partition_V1 and vaccine
  • Add citations for pre-packaged models partition_V1 and vaccine to paper.bib
• Alyssa in acknowledgements
• Gleamviz section
• Schedule 1 on 1 with Lauren to discuss authorship

[ancantu]

As a reader not entirely privy to epidemiological modeling:

episimlab/paper.md

Line 87 in ce4b880

`epydemic` is a Python package that provides a common framework for building models of epidemic processes [@epydemic]. It supports simulations that are discrete-time synchronous or continuous-time stochastic (Gillespie). Like Episimlab, it supports a generic model for compartmental disease, allowing for flexibitlity in the compartmental model structure. In addition, it ships with several basic compartmental models such as SIR, SIS, and SEIR.

SIR (mentioned in other lines as well), SIS, and SEIR are mentioned without any explanation for what the acronyms stand for. For an audience that is not familiar with these models, it might be helpful to write out what they stand for.

episimlab/paper.md

Line 58 in ce4b880

Inspired by previous works, Episimlab aims to support development of models with arbitrary complexity.

The explanation prior to this line about how other packages have limitations reasonably demonstrates why there is a need for Episimlab. However, I feel that this line (line 58) does not accurately reflect that your SDK will make up for the limitations present in other packages. I would reword to make this more clear to a general audience.

Overall, I could reasonably follow along for a substantial amount of it, and I believe addressing the comments above will help for a general audience.

[ethho]

@kellypierce I'm looking around on the consortium publications page for the citations for partiton_v1 and vaccine models... Are these two from Suzanna the ones we'd want to cite?

• https://doi.org/10.15781/cytm-ev55
• https://doi.org/10.15781/503k-c813

[kellypierce]

@eho-tacc here are some citations:

This manuscript is the first peer-reviewed article that describes the model implemented in partition_v: https://www.nature.com/articles/s41467-021-23989-x

This report is the first report that describes the vaccination model implemented in vaccine (nothing peer-reviewed yet):
https://doi.org/10.15781/gc9z-9h56

There may be other reports and papers we can cite as well; we can talk to other folks in the lab about how they'd prefer these works to be cited.