In the Ant module, we randomly selected 100 cells from the dataset and extracted gene set scores for each of these cells. We also retrieved a list of known human transcription factors (TFs) and identified intersecting genes between this list and the dataset. For each gene set score, we calculated the correlation and p-value between the gene set and different TFs using the Pearson correlation function (pearsonr). Any correlations resulting in NA were filtered out, and the false discovery rate (FDR) was adjusted using the fdr_bh method. Finally, we selected transcription factors with adjusted p-values below 0.05 and defined correlations greater than 0 as positive associations with the gene set, and correlations less than 0 as negative associations. The integration process for ANT begins by iterating through all BSID subdirectories under the BSID path. For each subdirectory, we check for the presence of an ANT result file named ANT.csv and read its contents using the pandas read_csv function. For each ANT result that is read, a new BSID column is added to record the current BSID associated with the processed data. Subsequently, all the processed ANT data frames are concatenated row-wise using pandas concat function, resulting in a comprehensive integrated ANT table (Supplementary Table XXX) containing the combined results.