Content of review 1, reviewed on October 13, 2023

The authors present a workflow to automatically extract quantitative trait information from a large number of images (termed: imageomics).
They outline the generic steps and provide computational components to perform them.
Additionally, a case study using, adjusting, and extending this workflow is presented.
The described workflow and published components are a valuable resource for research groups aiming to perform imageomics projects.
Many best practices for reproducibility and Open Science are integrated.

My two main concerns are
1. the transferability of the workflow components
2. the readability of the manuscript

Regarding 1:
I tried to follow the instructions in the REAMDE of the Minnow_Segmented_Traits repository to run the pipeline locally. I could not proceed very far, as I did not request a Fish-AIR API key, in time. Still, some observations:
- the R installation does probably not work as intended (the "Library" folder is created via mkdir in the README and again in the R script as "/Library" which expectedly produced a permission denied error, it is anyway not used further in the script and after installation the "Library" folder is still empty)
- some parts are still hard-coded, e.g. the SLURM account and the conda environment name
- the README claims, that docker is required, which is rarely present in shared HPC environments. The script seems to use singularity instead, which is more likely to be present.
- I did not perform a proper review of the source code. Given that the code is the main research product presented here, this might be a good thing to do. However, this is hardly possible through the anonymized link rather than the real repository.

Regarding 2:
I found it a bit hard to follow the manuscript. While this comment is very general, I try to give some specific comments that might help improve readability:
- for me, the story line is unclear. Do you present (a) the generic workflow with specific implementation to demonstrate utility, or (b) the specific workflow with some generic components extracted for easier re-use?
- somehow treatment of the general workflow and the case study are mixed, a clear separation of these two in the manuscript could help the reader
- in the case study, the aim is not clearly stated. I can't tell which traits are extracted (what are trait blobs? referred to multiple times, e.g. in l. 289)
- the structuring in small sections is a good idea in general, but I got confused what is a section/subsection/subsubsection.
- there are multiple personas (component creators, software engineers, and domain scientists) that are referred to multiple times in the manuscript without being properly introduced/defined
- heavy referencing of supplementary material; at least Table S1 should be included in the main text

Minor comments:
- l.132 one of the arguments for snakemake is that python is common in the field of imageomics, but in l. 165, the setup of the environment is only about R, not python
- l.208 numbers are swapped, it should probably be 13 species and 273 images
- l.260 how were these dimensions determined to have minimal spatial distortion? Was this done specifically on this dataset? Was any kind of padding used?
- l.303 R version 3.1.0 is cited, while in the repository README R version 4.2.1 is loaded
- l.304 include versions of python and snakemake
- l.311 for the data to be free and open, goes beyond FAIR, in the FAIR principles, only the protocol needs to be "open, free, and universally implementable" (see also https://danielskatzblog.wordpress.com/2017/06/22/fair-is-not-fair-enough/)
- l.386 what is the key point of the "Workflow Automation" section? Is it recommended to start manually and move towards automation as necessary? Or are you advocating for automation from the start, building on your generic workflow?
- l.422 machine-readable citation information, e.g., in cff format (https://citation-file-format.github.io/)

Source

    © 2023 the Reviewer.

Content of review 2, reviewed on February 29, 2024

The authors responded to all concerns by the reviewers and revised the manuscript accordingly. The manuscript is massively improved. The source code has been reviewed by another reviewer. I have no remaining concerns.

Source

    © 2024 the Reviewer.

References

    A., B. M., John, B., M., M., Bahadir, A., Yasin, B., L., B. J. H., David, B., R., F. C., Jane, G., Anuj, K., Kevin, K., Paula, M., Joel, P., Dom, J., Thibault, T., Xiaojun, W., Hilmar, L. 2024. A FAIR and modular image-based workflow for knowledge discovery in the emerging field of imageomics. Methods in Ecology and Evolution.