Skip to content

Latest commit

 

History

History
70 lines (52 loc) · 3.08 KB

Steps to Add New Reference Data.md

File metadata and controls

70 lines (52 loc) · 3.08 KB
Raw

R Notebook

On Adding New Data to Our Reference

The workflow of our identification model can be adapted to include new datasets in our reference by the following process:

  1. Obtain well annotated kidney cell dataset in a Seurat Object, placing the original cell type labels in a metadata column named ‘Original_Cell_Types’, and the study the data is from in a metadata column named ‘Study’. Ideally, the new data will also have information for the other metadata columns (If you are reading this, I will assume you either know how to check the metadata of a Seurat object and add your own or will understand the process after this linked tutorial.

  2. Merge new data with the existing data using a workflow similar to the CreationOfMergedObject.ipynb

  3. Plot a heatmap of the correlations of the new annotation labels along with the existing ones in a workflow similar to the correlational heatmap creation in figures.md. This will show whether the new cell labels should be included in an existing category or are part of an entirely new one. We assigned our annotations to categories based on both the correlations between cell types and the descriptions of the cells inside. In our workflow, any added cells must undergo this same manual assignment process.

  4. Run IntegrateAll.R and QualityControl.py. These will not be affected by new data.

  5. Now, in IntegrateSome.R, there is a variable called ‘iterator’ at line 14. The name of the new dataset needs to be added, as this will be the name of the train-test split object produced in which the new data is the testing dataset.

    Additionally, there is a for loop that currently runs 1:5 at line 26, this needs to be changed to run for the new appropriate number of datasets.

  6. In PerformanceTesting.py, the variable ‘withoutnames’ needs to be updated with the appropriate names for testing dataset iterations.

    Next, the original labels from the new dataset need to be assigned to a the appropriate Leveled_Names in the rename() function found in lines 34-84.

    Then, update the iterator in lines 115 & 116 to an appropriate number to include the new data. You will also need to add a list to the nested list of lists in this cell.

    Next, add a list to the nested list of lists in line 175

    Then, add your new dataset Study label to the df.columns in line 193

    Next, add a list to the nested lists in line 223

    Then, update the range at line 242

    Next, add your study to ‘datasetlist’ at line 269

  7. Finally, create the new reference. Take the final seurat object from IntegrateAll.R, and remove the cells identified in SVM Outlier Detection. An example of this can be found in lines 7-17 of IntegrateSome.R

  8. This is now your new reference object, which can be used in place of the ref.RDS downloaded from Zenodo in our Colab Annotation workflow!