Open eleozzr opened 4 years ago
Hi, As demonstrated in the paper, the p-values inform less on the significance of the interactions than the LRscore. Therefore, the p-values are not computed. However you can consider the LRscores as significance information, i.e. higher LRscore means higher confidence and lower LRscore means lower confidence. The equivalent of the 5% threshold on the p-value is a 0.5 threshold on the LRscore. Thanks for using SingleCellSignalR!
SCA
Hi, As demonstrated in the paper, the p-values inform less on the significance of the interactions than the LRscore. Therefore, the p-values are not computed. However you can consider the LRscores as significance information, i.e. higher LRscore means higher confidence and lower LRscore means lower confidence. The equivalent of the 5% threshold on the p-value is a 0.5 threshold on the LRscore. Thanks for using SingleCellSignalR!
SCA
Hi, All datasets in your paper are related to immune cells. Although the LRscores can as significance information, does 0.5 is appropriate for other types of dataset? By the way, how do you obtain the p-value when you considered the p-value as the cut-off? Just random permutation of the cluster labels of individual cells?
Hi, From my personal experience, 0.5 is globally appropriate. You can raise the LRscore threshold to 0.7 and the tol argument (from 0.01 to 0.05) for difficult datasets or convenience. The p-values are calculated by random permutations of the cluster attribution vector of all cells.
SCA
Good day,
It is a great tool to detect the ligand-receptor pair. I have tried to apply it to my dataset, i want to see the pvalue for each ligand-receptor like cellphonedb, but I didn't find the related function to get the p-value. Could you post the related code? Thank you very much.