BioModel Discussion Issue for eos18ie

[Zainab-ik]

Summary !!!

• BioModel Name: Wong2024 - Discovery of a structural class of antibiotics with explainable deep learning

• BioModel Tag: fair AIML, Machine learning model, Ersilia

• Metadata Comments:

  1. Does this model have multiple outputs? the paper is pointing towards that.
  2. the paper discussed models for predicting cytotoxicity to HepG2 cells, HSkMCs or IMR-90 cells based on the compound’s chemical structure. this is a separate models from the antibiotics, are the two models incorporated and do i annotate based on both or just antibiotics (the repository specifically stated antibiotics).
  3. Is there a likelihood of not performing a cross validation when building a model? there was no mention of it in the paper.

• Contributor:

  roles	Names
  Curator	@Zainab-ik
  Code contributor	@Richiio
  Reviewer	@GemmaTuron

• Curation status : manually curated

• Annotation file: here

[GemmaTuron]

Does this model have multiple outputs? the paper is pointing towards that. Yes, the original paper describes antibiotic potential and low cytotoxicity, but we are only implementing the antibiotic potential? Upon revising the code, it seems there are more checkpoints than needed in the model? outside the final_checkpoints repo the paper discussed models for predicting cytotoxicity to HepG2 cells, HSkMCs or IMR-90 cells based on the compound’s chemical structure. this is a separate models from the antibiotics, are the two models incorporated and do i annotate based on both or just antibiotics (the repository specifically stated antibiotics). It is separate from the antibiotics indeed Is there a likelihood of not performing a cross validation when building a model? there was no mention of it in the paper. In figure 1 I see a cross-validation if I am not wrong

[GemmaTuron]

Hi @Zainab-ik

What does this stand for? bqbiol:model | hasProperty | STATO: the statistical methods ontology | https://identifiers.org/STATO:0000277 | Hit

[Zainab-ik]

Fig 1b shows cross-validation and included in metadata. The Hit is a metadata to represent identification of hit compounds in screening.

[GemmaTuron]

Hi @Zainab-ik Let's not use Hit as it is very general, the rest looks good