Did you also remove the corresponding Q-score value? If you used trimmomatic/bbduk they would do this but if you used some other method then you may have an extra Q-score value remaining on line 4 for each fastq record.
I am not sure why you are having a problem but if I use the two reads you posted above I am able to demultiplex them fine (after taking the N's out).
I am not sure why you are having a problem but if I use the two reads you posted above I am able to demultiplex them fine (after taking the N's out).
Comment