MarinaBioLab Validation — Scientist walkthrough
Synthetic training example

0:00 — MarinaBioLab qPCR Validation
Hi, and welcome to the MarinaBioLab qPCR validation workbench. This tool turns a raw instrument export into a finished validation report. In about twelve minutes, I'll walk you through one complete validation, start to finish. You'll see where to click, what to check, and what to do when something looks wrong. Everything on screen is made-up training data, and the video has chapters, so you can jump back to any step later.

0:29 — Signing in
Let's sign in. There's no password. You use your work email instead. Type your address into the Work email box, then press Send sign-in link. Open that email on the same computer you're using now, and click the link. If you open it on your phone instead, the link signs in your phone, so type the eight digit code from the email into this page. And if no email arrives, ask your validation lead.

0:57 — Getting oriented
You're on the Dashboard. Let's get our bearings. The menu on the left is arranged in the order you'll work. Workspace, then Set up, then Data, then Output, and the AI co-pilot at the end. The strip across the top is called Your validation workflow. It shows the five stages of every validation, each with a short status and a suggested next action. Whenever you're unsure what to do next, read that text.

1:29 — Creating a session
Every validation lives inside a session. Think of it as a folder. It holds the panels you're validating, the study types, every file you upload, and everything the workbench produces. One session can hold several panels, so a multiplex validation stays together. We're on the Sessions page. Press Start a validation. After a moment the status says Session ready, and the header at the top shows which session you're in.

2:00 — Recording what you intend to do
Next, the plan. This is where you say what you intend to validate, before any numbers come in. You're on the Plan page. Tick the panels you're validating, then tick the study types you're running. Then press Build guided plan. If your lab already has a saved plan file, press Import plan JSON instead. Getting this right now decides how your results are organised later.

2:27 — Answering the open questions
Sometimes the workbench needs a decision that only a scientist can make. Then it shows a decision card like this one. Read the question and the reason it's asked. Pick the option that matches your evidence. If none fit and there's an Other choice, describe your own answer. Add the short explanation it asks for, then press Apply answer. It's saved with your name and the time. Clear every open card before moving on.

2:57 — Exporting from the instrument
Now your data. You're on the Upload page. One important detail. The workbench can't read a Bio-Rad dot p c r d run file directly. Export it first. In CFX Maestro, open your run, go to Export, then Export All Data Sheets, and choose CSV. QuantStudio exports, R D M L files, and Excel workbooks work too. If you pick the wrong file type, the message on screen reminds you what to export.

3:31 — Put the study in the file name
Here's the habit that saves the most time. Put the study type in the file name. Look at the selected file. Its name includes the word Accuracy, so the workbench can route those rows to the Accuracy study on its own. If the name doesn't say which study it belongs to, that's okay, but you'll confirm it by hand, one group at a time, in the next step.

3:55 — Save and queue the parser
Next to the file, fill in the Reagent kit catalog and the Reagent lot number. They're recorded with each upload for lot tracking, along with your name, and they never appear in a client report. Then press Save and queue parser. This is the step people miss. Choosing a file doesn't read it. This button does. Open Jobs in the header, and when the parser says succeeded, you're ready for mapping review.

4:24 — How the rows were grouped
You're on the Mapping review page. This step decides what your numbers mean, so slow down here. Press Load mapper groups. The workbench sorts the rows from your file into groups, one per sample. Each table row is one group. You can see the file, the sample name, which study it went to, where that routing came from, how many source rows it covers, and a reason in plain words.

4:51 — Where the routing came from
Look at the Source column. It tells you how each group was routed. Nomenclature means it recognised the study from text in your file. Parser guess means the parser suggested a study role. Unresolved means it couldn't tell, so you decide. Positive controls and no template controls are tracked separately. For each group, check the sample, its source rows, and the reason before you accept the routing.

5:20 — Making a correction
Sometimes a target name in the file isn't quite right. That's what Manual target correction is for. Press Load server target choices. Pick the measurement to fix, choose the correct target from the list, and type a short scientific reason. Press Preview correction to see the before and after values. If it looks right, tick the confirmation box and press Confirm target correction. Every correction and its reason stays in the history.

5:51 — What starts the calculation
Now let's calculate the results. Once you confirm the mapping and the session is ready, the workbench usually queues the calculation for you. To check, open Jobs from the header. If nothing has started, go to the Dashboard and press Queue next output job. If something is blocking it, the message says exactly what's missing. You don't have to wait here. The Results page shows the saved run when it's ready.

6:18 — The headline numbers
You're on the Results page. Across the top are the headline numbers. Overall status, mean detection rate, mean R squared, the largest C q gap, and how many targets and rows were used. Be careful with one of them. Mean detection rate averages the rates from each study summary row. It's not one percentage from every well. Treat these as a summary, not a verdict. The real story is in the study tables below.

6:48 — Reading a study table
Results has a tab for each study type. Each table has one row per metric. You'll see the study, the metric, the observed value, the criterion it's judged against, the status, and the source rows behind it. C q values show two decimal places. If a study was planned but has no data yet, its row says Needs review, Blocked, or N A instead of a number. The workbench never fills in a number it can't back up.

7:20 — The evidence behind a row
Want the evidence behind a result? Open the criteria review table. Each row names the study, the metric, the observed value, the criterion, and the status. The Detail column explains how it was judged. The Source rows column points back to the measurements it came from. Scroll sideways if the table is wide. Before you sign off anything marked for review, follow those references back to the saved measurements.

7:49 — Tracing a number back
Let's trace one number back to the raw data. Go to the Dashboard and find the Workbook explorer. Pick a sheet, then pick a row, and its details open. Every cell shows the formula that produced it and the source rows it used. Source row comparison puts the calculated value next to the original measurements. Explain formula describes how the detection rate is worked out. It's a reminder, not a substitute for checking the sources yourself.

8:20 — Figures and dynamic range
The figures come from the same result run. Target detection rate, mean C q by target, and the standard curves. Where a panel reports a semi quantitative range, the bands come from the fitted curve for that target rather than a fixed table, and a fit too weak to separate the bands is reported qualitatively instead. If a figure looks wrong, the repair is upstream in the data or the mapping.

8:47 — Checking the plate
Now let's check the plate. You're on the Plate map page. It's built from the calculated results, so it appears after the calculation has run. Use it to confirm your controls and replicates landed where you expected. Each well shows its call. Click any well to see the sample, the target, the C q value, and the source rows behind it. If a sample ended up in the wrong well, you'll spot it here.

9:16 — The part only a scientist does
Now the part only a scientist can do. Accepting the result. You'll find Result-run review on the Dashboard and on the Results page. Once you've checked the evidence, pick a rationale code, write your scientific rationale, and press Approve result. That puts your name against this exact result package. For a known deviation, record an Approved deviation, or file an exception with the target, the study, your summary, and the source rows. Accepting a result is not the same as approving a report. That comes later.

9:56 — Audience and branding
The report is built from the results you just accepted. Before you generate it, choose who it's for. Scientist review gives the full detail, with every table and source reference. Client final keeps the study conclusions and the key performance results, including precision standard deviations and CVs, and trims internal detail. Both use exactly the same calculations. Then choose Marina brand or White label. Choosing Client final describes the audience. It doesn't approve the report.

10:32 — Edit, or regenerate
You're on the Report page. There are two kinds of fix here, and it matters which one you use. If the wording of a section is wrong, press Edit section, change it, and press Save section. That's what a draft is for. The Linked to this section panel keeps its tables and figures attached. If a number or a table is wrong, don't edit the report. Fix the data or the mapping upstream, generate the results again, and the report rebuilds itself.

11:04 — Draft until somebody reviews it
This strip shows where the report is in its life. Preview, QA, independent review, then Approved. Right now it's a Draft. Every draft download has a DRAFT watermark on each page, so please don't send drafts to clients. To become Approved, the file passes its integrity check, the governing policy is ratified, the release checks pass, and an independent reviewer approves that exact version. Until then, Approve certified report stays disabled, and the list underneath says what's missing. The automatic QA pass only gives advice. It can't approve or reject anything.

11:48 — Draft versus approved
You're on the Downloads page, where every finished file lives. Each file shows its type, the result it came from, and a short fingerprint so you can tell versions apart. Before you send anything, read the button. Internal preview means on screen only. Awaiting approval means the report isn't approved yet. On the Report page, the same file reads Download DRAFT watermarked PDF until approval, and Download certified report PDF after it. If it says draft, it's not for a client.

12:26 — The reference definitions
Let's look at the Panel registry. It's the reference library the workbench uses. Pick a panel and you'll see its targets, each with its channel or dye, its oligo mix, and its LoD floor. The panel row shows how many study types it supports and its default C q cutoff. The key word is default. Whether a well counts as detected also depends on the amplification status and the run settings. Every session records the registry version it used.

13:00 — Asking it well
Now the AI co-pilot. Open it from the header while your session is selected. A good first question is, what evidence do we have for this session, or which review steps are still open. Wait for the full reply. Then, and this matters, check its references and the values it quotes against the tables. A reply about the whole session doesn't tell you the result for one target or study. Ask about the specific thing you need.

13:30 — What to hold onto
Treat every co-pilot answer as a draft for you to review, not a conclusion. Compare what it cites with the saved tables and source rows. If it doesn't answer your question, ask a narrower one, or use Discuss this value on the exact result cell. And remember, the co-pilot can't fill in missing measurements, make study decisions, or approve anything. Those still happen in the validation record, by you.

14:01 — Where to send it
What if you get stuck? First, read the written status on the step you're on. It names the blocker and suggests the next action. If something is taking a long time, open Jobs in the header to see what's running and whether anything failed. And if a number looks wrong, send your validation lead three things. The session name from the header, the file name, and what you expected to see instead.

14:27 — That's the loop
And that's the whole loop. Start a validation. Set your plan and answer the decision cards. Upload your file with the study type in the name. Review the mapping. Generate and approve the results. Check the plate map. Draft the report, get it reviewed, and download the approved copy. Come back to any chapter whenever you need it. Thanks for watching, and if anything looks wrong, please send it through.