Figures v3

[sreichl]

no action required, I am just collecting feedback/observations for now

• collect feedback
  • SCP4
  • Conference
  • Talk at HMS
• Discuss & Decide what to reject, what to include in preprint and what to postpone for paper v2

Feedback/Observations

• F1
  • B Wondering if we shouldn't move Modules way up and then have more whitespace instead of first whitespace. Reader needs guidance top left.
  • B capitalize Snakemake
  • C s of standard workflows is cutoff in PDF
  • D align end of Recipe box with collection arrow
  • Labeling: A is Modules, B is Recipes, C is projects, and D will be USP bottom right
  • try: move collecting arrows up such that arrowhead is aligned with downstream box and thereby Modules/Recipes/Projects get more top aligned with box and the Figure caption letter (A/B/C)
  • clarify
    • Modules/Recipes: Example: ATAC-seq quantification and Example: ATAC-seq analysis
    • Projects: Case study: Regulation of (human) hematopoiesis
  • bottom right (USP)
    • Should emphasize or make very clear that all subsequently shown analyses and results are done without a single line of code. Only "last mile" code is required e.g., for custom code (cross-prediction) or visualizations.
    • Positioning (here or at the top?) why MrBiomics vs NF-core/Nextflow/Galaxy: unique downstream modularization; assay-agnostic reuse, etc.
  • Benchmarks (teaser)
  • list of all modules connected with funnel from title Example: ATAC-seq quantification (top right )
  • list of all recipes connected with funnel from title Example: ATAC-seq analysis (top right )
  • move MrBiomics logo to the top left (ie start of the Figure)
  • rename NGS processing to NGS post-processing or NGS normalization & integration (imprecise but more descriptive)
  • consider flipping the whole figure ie start with projects and end with modules -> top down instead of bottom up
• Clarify what the exact output is (Which tables and figures will you get? Are the nice figures that you show there actual output of the recipe/modules): We claim MrBiomics modules produce figure-ready results (ie tables) and lots of visualizations but obviously not as pretty and fine tuned as we show in the paper as we consider this the "last mile"/custom 20% of every project that it is out-of-scope for us. Differently put: We want to get the user as fast as possible to the point of unknown and to answer their questions and take the tedious boilerplate engineering out of the equation.
• F3
  • A Identify divergent genes... sentence does not make sense, maybe instead "Identify divergent genes as remaining modality differences"?
  • C & E quite empty, looks off (not urgent). We could add (back) more PCs and metadata but distracts from main point for the sake of removing whitespace
  • I & J: colons missing in title after the respective cell types, otherwise we have to continue TA/EP with small caps.
  • I & J: remove bolding as most of the genes do not exhibit a phenotype in F4. let's stay figure inherent.
• Add figures to the Snakemake report probably as category Figures and subcategory either per dataset (as the folders are organized e.g., CorcesRNA) or per figure (e.g., Figure1). The metadata columns should probably either contain figure number or dataset (depending on the subcategory), type of plot (e.g., UMAP, CFA,...), etc.

USP/Portfolio info
F1 and F5 Should contain info on current offering/portfolio ie

• 11 modules
• 5 end to end recipes
• including what they covered

Current idea

• F1 just as statistic
• F5 with details ie modules names and recipe assays clearly stated, nothing too fancy mostly copying and labeling boxes in the same style we are using in all schematics throughout the figures

Two things:

1. I think it's fine if we repeat ourselves at the end of F1. To explicitly state what the unique selling points are. Somewhat as a summary and really high-level, what are the key takeaways of this whole figure to make it not implicit but extremely explicit. ie collecting the bottom lines/explicit characteristics across the whole (complex/comprehensive) figure in one place as list such that the reader does not have to think for themselves what all these things imply.
2. I think having the statistics (see above) and that it's fast & reproducible, and human & AI configurable is already good