This question is primarily directed at next-generation sequencing pipelines/workflows, specifically aimed at variant discovery. I was wondering if there are any additional practices, quality control steps, that are implemented in a clinical diagnostic setting which are not routinely used by bioinformaticians woking on academic research projects?
In a clinical setting, variant calls are being used for diagnostic purposes. Effectively they have to be correct, there is a lot less margin for error, and the pathological significance of variants needs to be relatively well supported. There are regulatory guidelines for clinicial bioinformatics as well governed under CAP and CLIA in the US. Typically, this includes documentation of pipeline development, and validation of the pipeline against deeply sequenced in house samples and "gold standard" datasets (e.g., Genome in a Bottle), so as to assess pipeline sensitive and specificity for known variants.
Couple useful links:
mforde84 has summed it up well, CLIA and CAP guidelines need to be followed for clinical NGS. Some additional resources for you:
In clinical diagnostic labs, one important QC distinction is that controls with established known positive variants must be included on every sequencing run that includes one or more patient samples. If sufficient coverage and variant frequency for these variants can't be confirmed, the run must be failed. The thresholds for these values are established by limit of detection analysis. Well characterized wildtype/negative controls must also be present to confirm against contamination with false positives.
Of course academic labs and genomics cores sometimes include positive controls, but is not a strict requirement, and they may not have to throw out all of the results if QC criteria aren't met.
Pretty much as mford84 - for the UK labs there are some guidelines available from the ACGS website
Inspections are usually via UKAS https://www.ukas.com/ (including for NGS tests and associated bioinformatics pipelines).
There's been a fair amount of additional work on how to validate against gold standards such as Genome in a Bottle, and a growing body of control samples to test against (mainly an Indel specific dataset atm). Most labs also have large in-house control datasets derived from Sanger sequencing and Microarray data - although this of course only tests for variants that have been found by these technologies previously.
Specific points for the NGS pipelines are audits, version control via something like Github, revalidation on software updates etc.
Edit: Should also add that at the moment many UK labs will also Sanger confirm any likely pathogenic variants (ddPCR or MicroArray for CNVs), found by the NGS pipelines.