If you’re relying on remnant patient samples to tell you how well your lab's bioinformatics pipeline can call clinically important variants, you might be missing more than you realize. In our experience, the bioinformatics pipeline can be the weakest link in assay development for many labs. Just because a variant is sequenced correctly doesn’t always mean that it will be called. And false-positives are just as bad. Sometimes it’s an issue of allele frequency. For example, we’ve seen cases where labs could detect certain mutations at 10% allele frequency, but as soon as the frequency dropped to 7%, they stopped detecting it. Other cases are caused by the complexity of the variant. For example, even at low allele frequencies, a lab may pick up relatively easy-to-detect single-nucleotide variants (SNVs) but can have problems with insertion/deletion (INDEL) calling errors. In both examples, the mutations aren’t missed because of sequencing or library preparation problems. As we’ve witnessed time and time again, when labs optimize their bioinformatics pipelines, they start picking up the low-frequency and difficult-to-detect variants again. The catch is, you first have to know you’re missing something. In assay development, what you don’t know can seriously weaken your test.