Warning! Not convergent! - Step 4: Measuring baselines

[wenkho]

Dear authors

Thank you for developing this package.

I am encountering errors on "step 4: measuring baselines ..." where the following warnings show up but copykat still continues to run and generate the outputs.

Could you please advise despite the above errors, the results generated by copykat is appropriate?

Thank you

[Evan-fnlt]

[1] "running copykat v1.0.5 updated 07/15/2021" [1] "step1: read and filter data ..." [1] "31492 genes, 43432 cells in raw data" [1] "10479 genes past LOW.DR filtering" [1] "step 2: annotations gene coordinates ..." [1] "start annotation ..." Warning: stack imbalance in 'vapply', 8 then 11 [1] "step 3: smoothing data with dlm ..." [1] "step 4: measuring baselines ..." number of iterations= 72 number of iterations= 304 number of iterations= 195 number of iterations= 156 number of iterations= 88 number of iterations= 69 [1] "low confidence in classification" [1] "cell: 1" WARNING! NOT CONVERGENT! number of iterations= 500 [1] "cell: 2" WARNING! NOT CONVERGENT! number of iterations= 500 [1] "cell: 3" WARNING! NOT CONVERGENT! number of iterations= 500

I encountered the same problem

[abollol]

me too

[gaobio]

Yes, 'not convergent' indeed occur quite often. GMM modeling for individual cells is the last strategy that copykat tries to estimate normal cell baselines if clustering strategy doesn't work. For individual cells, it not always convergent to GMM model. But we don't have a better model yet.

[Larrycpan]

Yes, 'not convergent' indeed occur quite often. GMM modeling for individual cells is the last strategy that copykat tries to estimate normal cell baselines if clustering strategy doesn't work. For individual cells, it not always convergent to GMM model. But we don't have a better model yet.

Could Prof. Gao provide an opportunity for users to modify the interations the normalmixEM, which is set as default 500? Maybe this can resolve the problem