It is trivial to create an extremely sensitive somatic mutation detection method by identifying any site with a single non-reference read as a candidate mutation. Clearly, such an approach would have an enormous false positive rate. Therefore in evaluating the performance of a mutation detection method, it is critical to thoroughly characterize its specificity. There are two sources of false positives: (i) over-calling events in the tumor and (ii) under-calling true germline events in the matched normal. Over-calling in the tumor is typically due to sequencing errors and inaccurate read placements whereas under-calling of true germline events in the matched normal is often due to low sequencing depth in the normal.
