When can PCR duplication distort apparent allele counts?

Why

During library preparation, fragments may be amplified many times before sequencing. Ten reads originating from one original molecule do not provide ten independent observations of the underlying sample. Preferential amplification of an allele can inflate its apparent fraction, especially in low-input or deep targeted sequencing.

In conventional paired-end WGS, duplicate marking uses mapping positions and read-pair geometry to identify likely duplicate fragments, with context-specific caveats. In highly targeted or amplicon assays, independent molecules can share coordinates, so simplistic coordinate deduplication may remove true molecules. Unique molecular identifiers, if introduced before amplification and properly analyzed, can identify molecule families and support error correction.

Compare raw and deduplicated depth, UMI family counts, strand support and validation controls before setting a variant-detection limit. More reads do not necessarily mean more molecular evidence.