Reply to Sheward et al.: Two sampling issues in using Primer IDs

We thank Sheward et al. (1) for their letter regarding our article (2) and for reminding the reader of the importance of sampling depth when using unique sequence tags (Primer IDs) to determine the starting number of templates. A Primer ID reveals previously unseen biases and errors from PCR and deep sequencing protocols, especially when applied to highly heterogeneous viral populations. One of the major artifacts it overcomes is resampling of the PCR product.

Sheward et al. (1) bring attention to another sampling issue, which is the sampling of the Primer ID library. As they note, the level of Primer ID resampling can be calculated based on the size of the library and the number of unique sequence tags that appear in the final dataset of consensus sequences, with the problem being that resampling of the same Primer ID sequence tag will result in more than one template with the same Primer ID. However, an experimental feature of the protocol limits the impact of this sampling issue. In those cases where the same Primer ID sequence tags multiple templates, whichever sequence is resampled the most will be recorded as the consensus sequence. Additional templates with the same tag are lost in building the consensus sequence, much like the removal of PCR and sequencing errors. Thus, although Sheward et al. (1) correctly state the sampling issue, we believe its impact on the use of a Primer ID as we described it is minimal. However, users of the Primer ID approach should keep these issues in mind when the sequencing of larger numbers of templates is attempted, most simply by extending the length of the Primer ID such that the number of possible sequence combinations is significantly undersampled.