News

HUPOST: Where does your interest in quality control come from?

DT: I’ve been in bioinformatics for twenty years now, and I have come to realize that black boxes harm users as well as bioinformaticists. I have therefore started to dedicate quite a bit of time to bioinformatics education. Many researchers in proteomics and metabolomics simply do not know the extent to which technical variation may be hiding biological variation in their experiments, so quality control represents a look inside the black box of these important methods.

HUPOST: Can you tell us  a bit more about the HUPO PSI Quality Control Working Group?

DT: The Quality Control Working Group (QC WG) is the first new Working Group of the HUPO Proteomics Standards Initiative in over a decade. It joins long-standing Working Groups such as the Mass Spectrometry and Proteomics Informatics Working Groups, which are responsible for mzML and MzIdentML/mzTab, respectively.

DT: The QC WG is a group of researchers from proteomics and metabolomics that is dedicated to see quality control mature as a research discipline.  We must ensure that QC is accessible to people who may not be familiar with the intricacies of standard formats and statistical models. The focus of the QC WG, therefore, is on making quality control a part of everyday practice.

HUPOST: Who should be interested in the Quality Control Working Group?

DT: I strongly believe that quality control is of interest to everyone in the field. Just because something is not dramatic does not mean it’s non-essential. When I think about the ways in which our work can be useful, I think about a principal investigator writing a solid research proposal, about a core facility director who wants to deliver top-notch service, and about a technician besieged by people who want their data “yesterday.” All of them need quality control, and we hope to arm these people with useful tools. The focus is currently on quality control for mass spectrometry, but the scope within this field is very broad, and includes both proteomics and metabolomics applications.

HUPOST: What are the desired outcomes of the work of the Quality Control Working Group?

DT: The primary target in our sights is interoperability; if a researcher creates a new set of quality control metrics, we want the output of the tool that calculates these metrics to be compatible with a statistical tool that was originally built to analyze another set of QC metrics. The qcML standard format for quality control metrics is at the core of this endeavour. To put it another way, quality control information is the commodity that we want to trade among metric generation software, statistical frameworks and repositories, with qcML as the coinage.

DT: Practically, this will require tools that generate metrics (for instance, QuaMeter and OpenMS) to produce qcML as output on the one hand, and statistical frameworks that can read and digest qcML on the other hand.

DT: Moreover, the objective of these statistical models should be to make tangible decisions based on these data, such as flagging an experiment as an outlier. Here, reliability is key. If a given tool repeatedly informs a technician that something is out of whack, the tech will quickly learn to ignore it if there are too many false positives.

DT: And of course, we endeavour to lower thresholds for adoption of quality control in everyday practice; the less effort required to emplace quality control regimen, the more likely people are to put it into use.

HUPOST: What would you say to prospective participants in the Quality Control Working Group to raise their interest?

DT: First and foremost, like all PSI Working Groups, the QC WG is always open to participation! The group consists of people who are very invested in quality control, who frequently bounce ideas off each other.

DT: But more than this, the QC WG really does need your insights; not only to learn what users require, but also your insights into statistics to harvest conclusions and decisions from quality control data. We also need the input from vendors; ideally their instrument control software would produce qcML directly to support assessment of these workflows, and vendors have great insights into the workflows that are right around the corner for our field!

DT:If you’re interested in participating, there’s a monthly Google Hangouts call, a QC WG GitHub repository, and a mailing list. Of course, the QC WG’s web page is also a good place to start.

HUPOST: What is the Quality Control Working Group working on right now?

DT: When qcML was first published, it was built around the available knowledge at the time. We are now extending qcML to accommodate a broader range of experiment types (such as selected reaction monitoring and data-independent acquisition). To achieve this, we need to address two challenges: how do we broaden the scope of qcML into new types of experiments, and how prescriptive should the QC WG be with regards to the description of quality control metrics? At our most recent meeting in Beijing in April, we came to the conclusion that a developer needs to have access to concise reading materials that allows them to create a valid and useful qcML quickly. It should not be necessary to understand the HUPO PSI working method and development process to be able to write a functional qcML file.

DT: Both of these challenges again boil down to lowering thresholds to adoption: making sure that quality control can easily be applied to your experiments, regardless of their type, and enabling developers to quickly and easily adopt qcML in their software.

HUPOST: Does the Quality Control Working Group have a presence at the 2017 HUPO Conference in Dublin?

DT: Of course! There will be an overall PSI session at the conference, and we are always well represented in the Bioinformatics Hub as well.

HUPOST: Thank you very much, and we wish the Quality Control Working Group a bright future!