search

saeyslab / multinichenetr

MultiNicheNet: a flexible framework for differential cell-cell communication analysis from multi-sample multi-condition single-cell transcriptomics data
GNU General Public License v3.0
107 stars 14 forks source link

multi_nichenet_analysis Requires Writing All Combinations of Contrasts #78

Closed DarioS closed 1 month ago

DarioS commented 1 month ago

It should not be necessary to manually write all of the contrasts for the convenience wrapper function.

Perform a MultiNicheNet analysis in an all-vs-all setting.

contrasts_oi and contrast_tbl could be automatically generated based on factor levels. Also, since most people will be analysing human data, why is there no default value for lr_network? Ideally, lr_network = "human" would seamlessly fetch it from Zenodo.

maartenciers commented 1 month ago

It should not be necessary to manually write all of the contrasts for the convenience wrapper function.

In theory, it is possible to generate contrasts automatically based on factor levels, depending on the design complexity. However, the main drawback is that the output will heavily depend on the specific contrasts set, which directly impacts the biological question you are addressing. Automating this process would overlook the importance of carefully considering the questions you are trying to solve. Depending on the biological question, the appropriate contrasts can vary significantly, as not every experiment is designed simply to compare treated versus untreated conditions. MultiNicheNet was designed to tackle complex experimental designs. Therefore, it is crucial to manually define contrasts to ensure they align with your research goals. That said, out of curiosity, how would you automate it in such a way to handle more complex designs while easy to understand?

DarioS commented 1 month ago

Ah, it would not be suitable for more complex designs.