Quality Assurance/Quality Control in the U of M DNA Sequencing Core

We take great pride in our low error rate, and we go to great lengths to maintain it. These efforts include both proactive measures for the minimization of likely sources of errors, and post-analysis detection of likely instances of Core-induced mistakes. These measures are detailed here.

Quality Assurance: Procedures used to minimize the likelihood of errors.

  1. All samples are checked by a Sample Clerk for accurate labeling and similar client errors before we accept them for processing. That Clerk maintains custody of the samples in our collection until they are turned over to the technicians for processing.
  2. Computerized sample tracking ensures reliable flow control. Only the client can enter the sample-specific information at submission time. Data generated at the completion of a sequencing run is automatically and reliably associated with the correct sample tracking number, and returned to the correct client.
  3. Pipetting is performed with a worksheet detailing which sample goes with which primer. Technicians use handwritten coding systems to organize their samples for each set, and to minimize the likelihood of mis-pipetting.
  4. All templates are added to an upper corner of the well and visually verified before they are spun to the bottom.
  5. All primers are similarly added to the top, for visual verification.
  6. Premixes are prepared with all reaction components except template and primer so that the exact same cocktail is present in all wells (including standards).
  7. The final volume of the well is visually checked to verify the presence of the premix.
  8. The volume of the plate is scanned visually after the purification step (Seph columns) to verify consistency of volume, before they are loaded onto the sequencer.
  9. After the run, all results are scanned one-by-one to check for symptoms of problems or of common failures (premix error, sequencer malfunction, salt effect, etc.).
  10. All plates are pipetted in a rotationally-asymmetric pattern, and differing from plate to plate, to allow for reliable detection of plate swapping or plate rotation errors.
  11. If a standard fails or generates inappropriate sequence, then an automatic warning is generated, facilitating detection of plate rotation, swapping errors, mis-pipetting, premix errors, etc.

Quality Control: Measures implemented to detect problems after the fact.

  1. Each set of samples includes at least one standard with known purity and known performance (and often several). If any standards do not sequence properly, it suggests a possible problem.
  2. All samples are passed through a scanning system that scores them for read length and quality of basecalls. We can detect unusual decreases in our read lengths or our quality scores. We can also produce reports detailing the performance with respect to technician, sequencer, client, time/date interval, etc.
  3. All sample sets are scanned automatically to detect the presence of suspicious similarities between samples from adjacent capillaries. This detects any instance of capillary crosstalk.
  4. All results are run though an 'auto-comment' system, that can detect certain modes of failure. Although it is not reliable enough for generating the client feedback, this can be useful for grasping overall views of failure modes.
  5. Samples that fail under suspicious circumstances (singlet failures among larger sets of otherwise-successful samples; large-scale failures from otherwise-savvy clients; unrecognized failure modes) are repeated to test whether the sample was at fault, or whether processing may have been erroneous.
  6. Additional failed samples are repeated just for QC purposes, even if there is no indication of a problem in their handling. Typically, this is a spot-check of a few samples out of larger failed sets.
  7. Clients are offered a way to request repeats when they can't come up with an explanation for a failure. This is the last-resort detection of a problem in our processing. Client-requested repeats that identify improperly processed samples are the most direct metric of our errors.
  8. Core administrators frequently view global reports of read lengths, failure patterns, client success rates, etc. These sometimes suggest additional Q/C checks to be performed. Frequently these lead to discussion with the client to encourage additional diagnosis of any perceived problems (whether client- or core-derived problems).
  9. Occasional large-scale reassessments are conducted, during which the Director scans through a large block of samples to detect instances where the technicians' performances might be called into question.