Agenda and minutes for meetings

[lwjohnst86]

This issue is just to list agenda items and some minutes for each meeting. So each comment will be a meeting agenda and associated minutes.

[lwjohnst86]

These items can be edited in this comment (use the edit button on the top right)

---

August 24th meeting

Action items:

For @ClemensWittenbecher:

• Think about what functions end user will use
• ... or what input and output looks like (Describe what input and output should *exactly* look like #7)
• Work on Generate simulated dataset #6 mostly. This is a key first step.
• Think about the functions that will be used (Algorithm to establish internal and external functions #14)

For @lwjohnst86:

• Setup infrastructure for unit tests
• Complete my TODO items in Generate simulated dataset #6 (Update: mostly completely)

Discussion items:

• Deal with Generate simulated dataset #6
• Draft a workplan/roadmap for getting a workable/citable product (Update: work on this at later meeting)

---

Some meeting minutes:

• simulated datasets for each
  • study design: case control, prospective, etc
  • outcome type: time to event, binary
• Flow chart of input and output datasets, what goes in, what comes out.
• Testing framework:
  • Use a strictly defined (aka manually created and all numbers written manually) output to test the functions, but keep it small, since this would be time consuming
  • Use a more loosely defined (simulated with more random noise) output/input to easily test different numbers of metabolites, outcomes, sample size, and conditions
• List of functions in List of functions #18

(Idea: Compare using a data frame vs list vs other graph format for performance. Conversion into form usable by some graph plotting packages...)

• Action items for next meeting
  • Create a smaller, strictly defined output/input so we can start testing on it
  • Flesh out the list of functions more (List of functions #18)

[ClemensWittenbecher]

Next Meeting: Oct 5th, 2018

Action items
For @ClemensWittenbecher

• Draft a roadmap
  • Propose milestones
  • Propose timeline
  • Propose responsibilities and tasks for contributers
• Work on List of functions #18 List of functions
  • Structure into main_functions, helper_functions, getter_functions
• Simulate data to test the main_functions (and associated helpers and getters)
  • Define properties and format of the simulated datasets
  • Identify possible functions, packages that can be used to generate the datasets

---

Meeting minutes

(added by @lwjohnst86)

• Timeline until a first good working version: ~6 months
• Goals by Jan 2019:
  • Have mostly the main functions created and fairly usable
  • Begin creating the testing and validating framework (doesn't have to be extensive)
• NetCoupler only uses Cox model right now... maybe try to generalise to other methods? (Generalise NetCoupler to use other estimation techniques? #19)
• Control for false discovery rate (Does NetCoupler control for false discovery rate in some way? #20)
• Omar mentioned exponential random graphs or something like that... we should check into that.
• The overall workflow is metabolomics data -> PC-alg -> select subset of variables or clusters -> run NetCoupler
• Discussed the functions (more in List of functions #18)
• We need a play dataset to work with

[lwjohnst86]

Thoughts/notes from 2018-12-20 meeting

• case-control design, but also other designs.
• include a progress bar? for runtime?
• include a limit (as an option?) for the number of nodes/variables to create a network on.
• do a unit test for make network function
• lintr into travis?

[lwjohnst86]

closing for now.

[lwjohnst86]

2021-02-18

Notes

• compare implementation of below (Clemens version) vs the current (my) version

  • From Clemens: I have checked in the ceramides project– the criteria for direct effect identification were i) P < 0.05 in the non-neighbor-adjusted model (rule1), ii) P < 0.1 in all neighbor-adjusted models (rule 2), iii) same sign of the beta in all the models (rule 3). I did not correct for multiple testing in the primary selection process.
    It would be interesting to apply the current NetCoupler-version to the Ceramides project with these same selection cutoffs and reproduce the results. In principle, it would be good to make these cutoffs' manual setting amendable as an option.

• As idea: provide argument for options as list.

• Action item: Determine journal and decide on whether it is method/theory vs application paper

• Surv can have three arguments

  • Or do it as a pre-processing (Surv or strata in mutate)

• things throw off when adjusting for variables with similar names (e.g "cer" when there is a "cer14.0" in the network)

  • based on some pattern selection?

• coxph cluster() within the formula interface? Same with strata (but that could be put in preprocessing)

  • cluster doesn't work.

[lwjohnst86]

2021-03-08

Agenda

• Update on previous meeting items
• Discuss about useR conference abstract (Luke should write something and send out before this meeting)
• Look at code again for Adjusting has some pattern selection problems #158

Minutes

• DE adjustment doesn't work perfectly (bug). Add as issue
• We use original implementation, not the current. Make issue about it.
  • (can try other approaches later, like AIC)
• Figure out how to add cluster based on model_arg_list. Make issue.
• Next meeting after hearing from useR

[lwjohnst86]

2022-01-08

Agenda

• Introductions
• Timelines, todo lists (listed more here: Task list and discussion items for initial version release #164)
• Who's interested in doing what
• Use regular meetings to move project forward?

Minutes

• Do a co-working session (on Discord) Monday and Tuesday. Set up an task list issue for coworking session, with aim to have something tidied for publication of paper
• Daniel I. send project application so we can focus our work
• Eventually others will be "beta testers" on their projects
• Work in parallel on a methods paper as an outline (Daniel W. + I.) (Luke make issue to brainstorm), e.g. to add more explicit assumptions to the method, what datasets and questions that it could reasonably be used for
• Make some "real-world" simulated data to use for examples and for methods paper

[lwjohnst86]

2022-02-14: Co-working

(co-working in Discord)

Tasks:

• Add brainstorming thoughts to Brainstorm methods paper #173 for methods paper (Daniel W. and I.)
• Go through website and identify areas of improvement for reading and learning how to use NetCoupler
• Make a better real-world simulated dataset (or give ideas for it, not high priority)

[lwjohnst86]

2022-02-23

Agenda

• Updates on activities

Minutes

• Almost ready for submission to CRAN, need to go through Cox model and other bugs and than ready
• For methods paper, using different forms of the data like more metabolites, more noise, etc to test how well NetCoupler.
• For simulation/testing with real world, could we use a published dataset?
  • Any that we could use for a "positive control"?
• Fabian will test out package in Cox model soon

[lwjohnst86]

2022-03-31

Agenda

• Update on activities.

Minutes

• Decide by next week on journal for application paper, and the requirement for state and maturity of package. (discuss here: Brainstorm publication outlet and type #163)
• Luke will finish last issue than submit to CRAN.
• During next meeting, we plan how to do the writing.

[lwjohnst86]

2022-04-29

Agenda

• Update on previous items
• Discuss writing plan
• Discuss Fix cph #188

Minutes

• Let's do JOSS for package paper and think about the methods paper in the future
• Plan an in-person for summer time? Maybe Clemens come to Aarhus?