Platemaps aren't hard but everything around them is

So, I have this pet theory that you can bucket all biotech startups into one of two categories: The ones that have problems with plate maps and the ones that have problems with sample sheets. The idea is that the startups that use plates have a hard time communicating what was in those plates to their data teams (That’s metadata!) and the startups that work with samples instead of plates have a similarly hard time communicating what was in the samples. Now, this is obviously an oversimplification since some startups have problems with both. But if there are startups out there that have either one completely figured out, I haven’t met them.

On the other hand, most of us who have worked on these problems would agree that in theory, at least, it shouldn’t be that hard. Plate maps and sample sheets aren’t inherently complex. They’re just tables. We regularly work with much more complex data structures like relational databases and knowledge graphs. Plate maps and sample sheets should be a walk in the park.

And yet, they’re not.

The problem, I believe, is that while these tables are simple on their own, the processes and tools involved in creating them are far from it. In particular, the process starts long before a scientist walks into the lab. The reason there’s no canonical solution, no open source plate map editor that everyone uses, is because there isn’t a single, well-defined set of requirements that can support all the different forms of this long and ambiguous process.

For many experiments, particularly in early exploration, designing plates or samples starts out quite informally and can go through multiple iterations before they’re finalized. Hallway discussions and the occasional whiteboard meeting. Eventually, someone may decide to write something down.

But they’re not ready to open the ELN, let along a LIMS. So we all know what comes next: They pull up Excel. (Or maybe Google Sheets.) It has all the flexibility they need at this stage, they already know how to use it and it’s already installed on their laptop. The perfect solution… unless you’re the data scientist who has to track down and then parse whatever monstrosity they create. But building a solution that provides the same level of convenience and ease of use, fits in anywhere in this process AND provides an easily findable and parseable output… well, it’s not exactly easy.

For other experiments, the process is more standardized. If you’re running a high-throughput assay for the Nth time, you’re probably using a standardized template. The process is standardized to the point where you should be able to define a single solution with well defined requirements that will handle it. This is what LIMS are meant for. The requirements may be fairly complex, depending on the template, and there may not be an off-the-shelf LIMS that exactly fits your particular needs. But it’s at least doable, if expensive to build and maintain.

Still, getting to the point where you have this level of standardized process takes time and effort - both inside and outside the lab. Until you get there, and as long as your lab is doing exploratory work, you’re going to have plates and samples in that first category. Which means you need to find a way to manage them.

There are a number of startups these days working on problems in this space, and many of them have solutions for managing plate maps and sample sheets. I’m optimistic that better solutions will come out of all this. But in the mean time, the next time you’re wondering why you can’t just clone an open source plate map editor from Github and call it a day… well, this is why.

---

Scaling Biotech is brought to you by Merelogic - an independent consulting firm that helps early stage biotech startups get their data under control at every inflection point so they can focus on the science.