MD QC · Live system suitability

QC was a post-mortem. Now it runs live.

MD QC scores every system-suitability run minutes after the raw file lands on the instrument PC, against a gold standard your team sets. Drift shows up on the day it starts, not when someone finally opens the folder.

Open-source agent (Apache 2.0). Thermo · Sciex · Bruker · Agilent. Included at no charge with every Evosep Lupo and Evokit.

The MD QC Agent web UI: service state, upload queue, configured instruments and the most recent extractions
The agent’s own web UI on the instrument PC. Longer-range trends live in the Mass Dynamics platform.
Why now

Throughput went up. How often anyone looks at QC didn’t.

A weekly look at the QC was fine at 30 runs a day. At 200 and more, the same three days of unnoticed drift cost seven times the samples. Nobody can watch every QC injection. Something has to, on rules you set.

20–40
runs a day, a few years ago. Three days of drift before anyone looked: about 90 samples exposed.
200+
samples a day now. The same blind spot: about 600 samples acquired on a drifting instrument.
Minutes
from a QC file landing to a verdict with MD QC, so the day-four drift gets caught on day four.
What MD QC is

One product, in two places.

A small agent on the instrument PC does the measuring. The Mass Dynamics platform does the remembering: trends over months, across instruments and sites, shared with the people who need to see them.

01 · On the instrument PC

The MD QC Agent

A background service that watches your acquisition folder, waits until each raw file is fully written, and classifies it by filename: system-suitability control, process control, gold-standard run, blank or sample.

For every control it runs a real Skyline extraction against your own template, scores the run and writes one small JSON payload. Sample files are never touched.

02 · The reference

A gold standard you set

At install, or after a source clean or column change, you run 15–20 system-suitability injections per SPD and tick the ones that define the instrument at its best. Their median becomes the gold standard.

Every later run is scored as drift from that known-good state. Old gold standards are kept, dated and labeled, and every payload carries the ID of the one it was scored against.

03 · In the platform

The MD QC modules

With upload turned on, each payload lands in your Mass Dynamics workspace: a traffic-light scorecard per run, longitudinal tracking per peptide, and the same view for every instrument and site you connect.

QC runs sit next to your experiments, so instrument health is part of the same record as the biology it produced.

How a run gets scored

From raw file to verdict, every time a control lands.

No new method and nothing new in the acquisition path. The agent sits where the Evosep workflow ends: a raw file on the instrument PC.

01

Watch the folder

Any folder that receives raw files. Thermo, Sciex, Bruker and Agilent.

02

Wait for the file

Until it has stopped changing and the vendor lock is released, so a half-written file is never read.

03

Classify

By filename: SSC, PC, SSC-gold, blank or sample, with the samples-per-day alongside. Patterns are configurable.

04

Run SkylineCmd

Skyline’s own command line against your template, each run in its own isolated folder.

05

Parse the report

Retention time, peak area, mass error and dot products per peptide, from the report definition you choose.

06

Check the peak is real

RT within 2% of the gold standard means the right peak. If RT moved, area and dot product decide whether it is drift or a mis-pick.

07

Score and write JSON

Per-peptide ratios and deviations against the gold standard, with the thresholds applied, in one payload per run.

08

Show and share

In the agent’s web UI on the PC, and in your Mass Dynamics workspace if upload is on.

Every derived flag ships with the measurements and thresholds that produced it, so a verdict can be replayed, audited or re-derived under your own policy.

With Evosep

Evosep puts the QC in the plate. MD QC reads it.

Every Evokit ships with its controls built in and a fixed run order, so nobody has to design a QC strategy. Evosep standardizes what goes into the instrument. MD QC reads what comes out.

  • SSCSystem suitability controls check LC-MS performance over time. The first four injections of each plate are QC, so an experiment can fail fast; the rest are spread through the plate. MD QC scores each one as drift from the gold standard.
  • PCProcess controls check the whole sample-preparation workflow. In filenames they arrive as PCD96 or PCD384: one control, the number only records the kit size. MD QC tracks them as a ratio to the gold standard.
  • SSC-goldThe 15–20 runs per SPD, chosen at install, that define the gold standard every later control is read against.
Sample input, Evokit, Evosep Lupo, Evotip Pure, Evosep Eno, MS readout, MD QC, MD Discovery, biological knowledge
Sample-to-separation standardization by Evosep, extended by Mass Dynamics to reusable biological knowledge.
Announced at HUPO 2026: MD QC is included at no charge with every Evosep Lupo and Evokit, for as long as you hold the system. The agent plus hosted MD QC, with no limit on QC runs; plan details and the live QC-history window are as published on our pricing page. Read the announcement →
Set it up

Six steps on the instrument PC. About 20–30 minutes.

This is the short version of the operator guide. The full guide, with the complete config.toml, troubleshooting table and file locations, is in the repository.

1

Create your Mass Dynamics account Free · in the browser

The agent works as a local QC monitor on its own. Trends across months, instruments and sites need a workspace to land in, and the API token the agent uses comes from your account settings.

Create your account - free →   Already have one? Sign in

2

Install the agent PowerShell · Python 3.11+

One virtual environment, called by path so there is no execution-policy change to argue about with IT.

py -3 -m venv C:\MassDynamics\mdqc
C:\MassDynamics\mdqc\Scripts\pip install https://github.com/MassDynamics/md-qc/archive/refs/tags/v1.0.2.tar.gz
C:\MassDynamics\mdqc\Scripts\mdqc version
3

Put your Skyline method in place One folder

Copy your QC method (.sky), its spectral libraries (.blib) and the bundled MD_QC_Report.skyr into C:\ProgramData\MassDynamics\QC\methods\. The libraries must sit in the same folder as the method; Skyline resolves them relative to the document.

4

Configure and run config.toml · then one command

Create C:\ProgramData\MassDynamics\QC\config.toml from the template in the guide. The lines marked <-- EDIT are yours: the instrument label, the folder your instrument writes raw files to, and the vendor and file pattern if you are not on Thermo .raw. Save as plain UTF-8 without a BOM.

C:\MassDynamics\mdqc\Scripts\mdqc run --foreground

The startup line ends in service_started and prints the web UI link; the tray icon opens the same page. To connect the platform, open Settings → Cloud, paste the API token from your account details, save and restart. For an unattended install, register it as a Windows service (recipe in the repository).

5

Record a gold standard 15–20 runs per SPD

Run your SSC-gold set, open Gold standards in the web UI, tick the runs that look representative and save. The page shades any run or peptide that deviates from the ticked set, so an equilibration outlier is visible before you commit. Until a baseline exists, runs keep their measurements but carry no verdict.

6

Verify end to end About 90 seconds

Copy one raw file into the watched folder. After the stability window and the Skyline extraction, the dashboard shows the file, the targets found (for example 48/48), the mass error, and a new row on the scorecard. If targets read 0/48, the Skyline template does not match the acquisition; the troubleshooting table in the guide covers it.

Read the full operator guide →

Get going

Two ways to start.

Set it up yourself from the steps above, or tell us about your lab and the relevant people at Mass Dynamics and Evosep will help you get it running.

On your own

Create a free account, install the agent, record a gold standard. Setup is 20–30 minutes; the first verdict comes as soon as a gold standard is saved.

Create your account - free
Help set the thresholds

The default warn and fail bands are a vendor draft from one dilution series. If you want them set properly on your platform, run the SSC dilution series with us.

Book a working session →

Questions on an install: engineering@massdynamics.com, or open an issue on GitHub.

Running an Evosep system?

Tell us about your lab and your instruments, and we’ll be in touch to get MD QC set up.