The hidden step in your DOE journey ??
Hi there!
When you're working with customers, every day's a school day.
And recently, I discovered that?there's a hidden step in the Design of Experiments (DOE) journey... though it’s got nothing to do with the science.??
It comes up once you’ve run your DOE, analyzed your data, and got some interesting insights from it. At that point, it might be tempting to think that you’re done.?
But you’ve still got an important job to do. You’ve got to present your findings to your colleagues and leadership team. You’ve got to get their buy-in, by helping them understand how you’ve approached your experiment and why.
One of my customers stumbled at this unexpected hurdle recently. She’d just finished her first DOE.
She was impressed by how many conditions she’d been able to test, and she’d generated some really interesting data—so she’d presented her work to her team.??
But they didn’t appreciate the progress she’d made, because they were reluctant to infer too much from her analysis.
She’d tested a subset of conditions, and used them to infer information about the system as a whole—which is common practice in DOE.
But frustratingly, they focused on all the conditions she hadn’t tested, rather than the ones she had.??
To some extent, it’s natural for scientists to question everything—the term “scientific rigor” exists for a reason! We want to understand things fully before we place too much trust in them.
Except that statistical modeling doesn’t need to work like that.
Think of a model as a black box: You put data in, and get predictions out. You don’t need to know what happens in between to get value out of it. The process can be opaque, and that's the perfect breeding ground for mistrust.??
What’s more, multi-dimensional modeling isn’t intuitive for our brains to understand. We can only visualize 3 dimensions at a time, and most experiments in drug discovery have more variables than that. And analyzing multiple factors in a single dataset can really mess with your head.
The truth is that biological systems are inherently complicated, and we need complex tools to understand them. When we’re presenting our DOE results, explaining the underlying methodology is just as important as explaining the scientific background.
But hearing my customer's story made me realize that this is easier said than done.
So, I'm planning to tap into our team's 20+ years' experience with running and adopting DOEs to create some resources.
If you've ever struggled to communicate the value of DOE, I'd love to know what bits you found particularly challenging—and what you'd wished you'd had to hand at the time. Pop me a DM on LinkedIn, and I'll see what I can do.
Speak to you soon,
?? Markus' love-hate relationship with DOE
You'll be relieved to hear that our CEO and Founder Markus takes no issue with Design of Experiments as a method (he's still an avid fan). But in his latest LinkedIn post, he rightly points out a few reasons why the term can be misleading.? ?
?? "You can't dispense 0.1uL in the lab"??
Our Customer Success Scientist Nathan Hardingham clearly heard this, and thought, "Challenge accepted." He used Formulatrix's Mantis to dispense 0.1uL across an entire 96-well plate—and filmed it for all to see. It's a precise and pleasing watch, we highly recommend.
领英推荐
?? Biologists vs statisticians??
JMP's Phil Kay and Markus finish off their first season of The Next Experiment podcast with a socially taboo topic: Why biologists and statisticians butt heads when working on biological experiments. Together, they unpack the root causes, and explore how they can find ways to work better together.?
?? Content we're loving?
Why we loved it:
Quality should never be an afterthought. Luckily, the Quality by Design (QbD) approach, where quality is embedded into the development process from the beginning, is already being used successfully in drug manufacturing. And it's becoming commonplace in preclinical assay development, too.
In this overview, authors showcase QbD using 3 case studies of different assay types, specifically an AlphaLISA assay, cAMP biosensor assay and an arrayed CRISPR screen. Excitingly, they opt for Design of Experiments (DOE) to characterize the design space of each assay.
And bonus: They use Synthace to help them translate their DOE designs into experimental executions with a range of automated liquid dispensers.?
Why we loved it:
If you’ve ever taken a tumble on a slippery rock in a stream, you’ve likely encountered a biofilm. Biofilms are communities of microrganisms that form a layer over surfaces, helping them survive against various environmental stressors.
But for medical devices like implants or equipment in industrial settings, they can be problematic, as getting rid of biofilms is no easy task.
But it looks like nature itself might offer a solution. This study has found that plants, as well as bacteria and some parasites, produce a metabolite, methylerythritol cyclodiphosphate (MEcPP), which can disrupt biofilm formation. Normally part of the methylerythritol phosphate (MEP) pathway,?MEcPP causes decreased fimbrae production—surface structures needed for biofilm formation and adhesion to a surface or other microorganisms.
So, no harsh chemicals needed for cleaning after all.
Why we loved it: When most people hear DNA damage, they think of sunburn. Though it’s definitely part of the story, this study identified a completely new culprit causing the immediate discomfort following sunburn, like pain, itching and blistering of the skin.
It reveals that it's the UV induced damage to the ribosomal RNA or messenger RNA (mRNA) templates activating the ribotoxic stress response (RSR) that's responsible for the underlying cell death and inflammation of the skin.
The key player orchestrating this whole mechanism is the ribosome-associated mitogen-activated protein 3 kinase ZAKα. ZAK knockout mice were protected against early skin inflammation after UVB irradiation—proving that it's RNA damage and not DNA damage that triggers the initial response to acute sun exposure.
About Synthace
Get faster, smarter insights from your R&D experiments. Designed by and for biologists, Synthace lets you design powerful experiments, run them in your lab, then automatically build structured data. No code necessary. Learn more.
Get the latest from our lab team
Scientists Luci and Emily keep you posted on DOE, lab automation, plus all things current and future-facing in life sciences R&D. No frills. Only the important bits.
Customer Success Manager at Synthace
2 周I think that’s sound advice in a lot of cases Phil Kay - when time is precious, people care about results, not how you got there. And to Winfried’s point, it really depends on your audience, as your scientific colleagues are more likely to be interested in the technicalities than leadership. In the early stages of implementing a new way of working, there’s a need to build trust, potentially amongst many different stakeholders. And in my view, the best way to build trust is to bring people along for the ride, and make sure that they understand what you’re doing and why.
Manager Customer Centric AI @Shell
2 周The issue you describe is a case of advanced data storytelling. I believe, Phil Kay , you are actually right: the results is the most important bit, and in most industries that will be where the interest ends. But if you are in a scientific community you need to also communicate, how you arrived at the conclusions and why you are sure of the validity. Also, in scientific context I would always have a slide listing the assumptions made, and which conditions were considered but excluded…
DOE & Data Analytics Evangelist | Nervously excited about Digital Future of Science, Engineering, R&D, Manufacturing | Medium-pace runner and road cyclist
4 周I often tell early career scientists that are staring out in industry that they need to focus on communicating the results. Don't spend a load of time talking about how complicated and difficult the work was. Nobody cares. I'm rethinking this advice now.