Online Classification

We provide online cell type prediction of user uploaded single-cell datasets. Based on Scibet, it takes around 5 seconds to classify 500 cells in the sample test file and generate reports using Chrome. Please follow the instructions below to upload and analyze your data.


Step 1: Choose one file to upload

We currently accept gene by cell .csv files with the first column being gene names and column names being user-defined sample indexes. The second row of the .csv file should be annotated cell types and can be left the same as the first row(column names). Expression profiles should be library size normalized(CPM, TPM, etc). Performance would be further improved if only informative genes are included. Please refer to our sample files for dataset formats.


   scibet_mouse_bladder_test.csv

Note: if you are using our test file 'scibet_mouse_bladder_test.csv', you can skip Gene conversion below, and in step2 use the default reference.


Gene conversion

If uploaded file and references are of different species, you can convert genes in the uploaded file to match the reference using the following choices:

Step 2: Choose a reference file

Use pre-calculated reference

Please scroll down and choose a scibet reference by simply clicking the 'choose as reference' button. You can also use the search function of the browser (ctrl+F) for key words to directly locate to your reference of interest. If you are performing cross species analysis, please ensure gene names are compatible with the reference species.


Or Use your own reference

Format of reference file goes the same as test files


   scibet_mouse_bladder_reference.csv

Reference:

Step 3: Click to submmit


Below will show a histogram of cell type composition & a downloadable table of prediction results.



We provide a series of trained scibet models so that users can start directly from the minimized probability matrix, instead of downloading and processing the raw expression matrix. Users can perform online classification directly or download and load the models using the Scibet package. The list of references is under continuous updates.




Results