How does Picard's MarkDuplicates handle unique molecular barcodes and PCR error?
I'm interested in using unique molecular barcodes to help distinguish what is PCR error in my samples. My current understanding of how MarkDuplicates chooses a "best-pair" is that it chooses this best pair based on a high sum of base quality scores. Sequences containing PCR errors can also have great base qualities. Has any additional logic been implemented to help distinguish reads containing PCR errors from the original template sequence when using unique molecular barcodes?