search

BiomedicalMachineLearning / stLearn

A novel machine learning pipeline to analyse spatial transcriptomics data
Other
190 stars 25 forks source link

preprocessing steps #244

Closed derekrusso closed 9 months ago

derekrusso commented 1 year ago

After reading all the tutorials, I think stLearn is a great tool for spatial downstream analysis ! But forgive me for asking silly questions (I am new to spatial omics analysis):

the datasets used in the tutorials all have preannotated information, either by deconvolution or label transfer. However, stlearn currently do not offer cell type annotation function. Is it recommended to first annotate spots (eg. by tangram or seurat)? By applying the above pipeline, there seems problems regarding the preprocessing steps. (as both stlearn and seurat offer preprocessing, normalization, and clustering).

Is there a 'best practice' for analyzing spatial data by stlearn?

duypham2108 commented 1 year ago

For any type of analysis, I suggest that you should keep the raw data (which might store in different layers of anndata object). Currently, we don't have a best practice for preprocessing. You can transfer the result from other tools/software to a new anndata with your data and perform stlearn with it. It will be required a bit of hacking to the anndata object.

We will try to improve the tutorial and codes to support users using results from other tools as soon as we can.