How a Sample Was Made Matters More Than the Sample
Workflow lineage records lab work as steps with inputs and outputs, so any result can be traced back to how its sample was made.
By Yaoyu Yang ·
When an assay result comes back off, the first thing most scientists want to know is how the sample behind it was made: which culture it came from, what was done to it, and who ran each step. Our team has built several generations of sample tracking systems, and the lesson that keeps coming back is that this history matters more than the sample record itself. So at Cypher AI we built the LIMS around this concept, with a data model we call workflow lineage.
WHAT IS WORKFLOW LINEAGE?
Workflow lineage is a way of recording lab work as a chain of steps, where each step has inputs and outputs, so every sample and data point links back to the steps that produced it. A step might be a transformation, an overnight culture, a purification or a plate read. Each step records what went in, what came out, who ran it and when.
Most sample tracking starts from the sample and its container: an ID, a location, a volume and sometimes a parent sample. That record tells you what a tube is and where it sits in the freezer. It usually says much less about how the material inside was made, which is the part you need when a result has to be explained.
WHY ISN'T SAMPLE TRACKING ENOUGH TO EXPLAIN A RESULT?
Sample tracking tells you what you have. Explaining a result takes the history of how it was made.
Take a protein expression run. The purified protein sits in a tube with an ID, a concentration and a location. Say its activity assay comes back at half the usual signal. To find out why, you need to know which colony the culture came from, what temperature the induction ran at, which column and buffer lot were used in the purification, and what changed since the last batch.
In many labs, answering those questions takes lineage reconstruction, or what we call mental backtracking. Someone opens ELN entries from three different weeks, a spreadsheet of culture conditions and an email about a buffer lot, then pieces the path back together from memory. That can take an afternoon when the person who ran the work is still around, and a lot longer when they've moved on.
The cost shows up across the field. In a 2016 Nature survey of about 1,500 researchers, more than 70% said they had tried and failed to reproduce another scientist's experiment, and more than half had failed to reproduce one of their own. When a result won't reproduce, one of the first places to look is how the materials were made, and in many labs that's the hardest thing to determine.
Records also fade over time. In a 2014 Current Biology study, researchers asked for the data behind 516 papers published 2 to 22 years earlier. The odds that a dataset still existed fell by 17% for every year since publication. Notes on how a sample was made often live in one person's spreadsheet or memory, so they're best captured while the work is happening.
HOW DOES A WORKFLOW-LINEAGE DATA MODEL WORK?
A workflow-lineage data model treats workflows and their steps as the core records, and samples and data attach to them as inputs and outputs. In the Cypher AI LIMS, each step records the samples and containers it used and the samples or measurements it produced. Lineage builds as the work is recorded, so nobody has to link records together afterward.
Back to the protein run example. The transformation step takes a plasmid and produces colonies, the culture step takes one colony and produces a culture, the induction step records its temperature and time, and the purification step takes the lysate and produces the purified protein. When the assay result is recorded on a well, it's already connected to each of those steps, so checking the induction temperature means opening the step it came from.
Records don't need to be perfect the first time. A scientist can go back to a step and add a measurement, a note, or who ran it and when, and every change is tracked. Volumes also draw down from the source sample as material is used, so each tube's record shows what's left.
Cypher AI is digital infrastructure for life-science R&D: an ELN, a LIMS, data analysis and a services Marketplace in one platform, with an AI agent that works across them. A scientist can describe what they did in the ELN, and the agent proposes the samples, containers and steps to record, which the scientist reviews and approves before anything is written. That keeps the lineage complete without a separate round of data entry after every step.
WHAT DOES WORKFLOW LINEAGE CHANGE FOR AN R&D TEAM?
For an R&D team, the biggest change is usually less time spent reconstructing work and rerunning experiments that already exist. We often see people rerun an experiment a colleague in the same company ran six months earlier, because nothing pointed them to it. Ten-person startups we've talked with run into the same thing.
Lineage also helps when work changes hands. A scientist who joins a project can open a sample and see the steps behind it, and a reviewer can trace a figure back to the runs that produced it. Structured lineage is also the kind of context AI tools need to give useful answers about lab data, since a model can only reason about the steps that were recorded.
FREQUENTLY ASKED QUESTIONS
How is workflow lineage different from sample genealogy? Sample genealogy links a sample to the parent samples it came from. Workflow lineage also keeps the steps in between, so you can see what was done to get from one sample to the next.
Can I bring existing sample records into Cypher AI? Yes. Sample lists can be imported from a CSV through the agent. Lineage for those samples starts with the first step recorded in Cypher AI.
Does workflow lineage replace the ELN? No. The ELN holds the account of the experiment, and the LIMS holds the samples, containers and steps. In Cypher AI they sit in one platform, and an ELN entry can link to the samples it used.
HOW CAN YOU TRY WORKFLOW LINEAGE ON YOUR OWN DATA?
A good test is a workflow your team already runs, ideally one where a result has been hard to trace back. We're happy to walk you through it, or give you access to try it yourself. Start with a free account at cypherbio.ai and book the 15-minute onboarding call from the home screen.
For R&D organizations that need their own cloud, SSO or a forward-deployed engineer working alongside the team, the Enterprise plan covers that, and you can book a conversation through the demo link on the site.