Causal Inference Sub Group 2025

AWG Projects AIML AWG Open Projects
82

Apologies for the late notice, we are cancelling today’s meeting due to the Memorial day holiday in the US.

1 Like
83

Dear team @AIMLawg

Someone from the AI/ML AWG shared these resources for creating phenotypes from plant images for machine learning / AI analyses. We can discuss these at our next meeting:

The LCPC Transform Precisely Measures Multiple Morphological Features of Corn Kernels [8 min]

Shape Genie Precisely Measures Multiple Morphological Features of Corn Kernels​

Quantifying Morphology of Fungal Spores [10 min]

Quantifying Morphology of Fungal Spores​

Standardized Way of Photographing Mushrooms for Shape Analysis [40 min]

Standardized Way of Photographing Mushrooms for Shape Analysis - MFL University 2023​

4 Likes
85

Hi @james.casaletto and all,

Here is the pre-print formally describing this open-source algorithm. The paper was just accepted for publication and the most recent version has been uploaded to BioRxiv: https://www.biorxiv.org/content/10.64898/2026.02.02.703425v3

Dave

2 Likes
86

Thank you so much Dave for sharing!

87

Here are my scribbled meeting notes from the June 22, 2026 meeting:

  • James, Anna, and Daniya’s RR9 paper (with CRISP and other ML) has been accepted for publication at NPJ Microgravity

  • Anna proposing to lead a course for students on causal inference at ETH

    • starts in September
    • possible projects include

      • lunar regolith (needs data augmentation)

      • compare dowhy with crisp

      • which aspects of geodynamics “cause” the things we observe on the surface of the planet

  • James suggested that as a group, we

    • continue pursuing individual projects
    • ask questions of the group on projects
    • recruit help from group for projects
    • one thing we all rally around and continue discussing is how to validate in-silico causal inference findings
      • does the literature support the findings?
      • multiple lines of evidence ==> more confidence in results (e.g. CRISP + ML + DGEA + …)

  • Daniya asked about the multiplicity of functions per gene and how to disentangle

    • one gene can show different characteristics
    • what are other phenotypes it can show

  • Discussed we need to discuss what to do with causal inference findings in terms of risk management, mitigation, and countermeasures.

  • Rachel to share results and questions at our next meeting (just got 2 datasets from borja)

3 Likes
88

Hi James @james.casaletto, For the in-silico validation, I think you could also use retinal studies from missions other than RR9. Then examine whether the same genes are consistently identified as causal across these studies using the different models, such as CRISP, that you have already used in your work. This might provide additional evidence for the robustness of your causal inference approach.
Sorry, but I won’t be able to join the meeting this month.

1 Like