Blog & News

View RSS Feed
Garbage in, Garbage out...and other myths

Garbage in, Garbage out...and other myths

Last week, I had the privilege of attending Lab of the Future at the iconic Beurs van Berlage in Amsterdam. This stately, somewhat nostalgic building is a nice juxtaposition to the many talks on cutting-edge advancements in science and technology, for which this conference is well-known. However there was one aphorism I heard repeated again and again that, by comparison, seemed almost as old-fashioned as the vaulted brickwork in the main hall. That aphorism was this: "Garbage In, Garbage Out" It is a pithy truism born in the software development world, and it captures a simple truth: the quality of what comes out of a system is directly related to the quality of what goes in. Put in “garbage,” and you’ll likely get “garbage” back. As general life advice it’s perfectly fine, but given recent advancements in AI, I suspect we are adhering to it too rigidly in the life sciences. If the presentations I saw are to be believed, one canno...
Read more on page 6 →

Better Science with Spooky Tech

In countless conversations about AI in the life sciences, a recurring theme emerges: the pressing need for vast, clean, "AI-ready" data. People often envision AI as massive prediction engines like AlphaFold or other drug discovery tools that rely on enormous databases of protein sequences and omics data. While these are undeniably important applications, it's unfortunate that other, less glamorous (yet equally critical) uses receive less attention in pharmaceutical sciences. One such area is synonym management and ontology harmonisation, where AI's semantic awareness holds immense potential. To explain why—and to illustrate the frustration of dealing with systems that don't understand semantics (and since it's Halloween)—let me share a story about a Halloween-themed 8-bit text parser adventure game from the '90s called Hugo's House of Horrors. If you're too young (or possibly too cool) to remember, this game casts you as Hugo, whose girlfriend Penelope has been kidnapped following an ill-advised babysitting job at an obviously haunted house. Your mission is to explore the house and rescue her by directing Hugo with arrow keys and simple commands like "open door" or "inspect mask." Along the way, you're met with various spooky puzzles and nefarious monsters that frustrate your efforts. More often than not, however, the most difficult task was overcoming the game's restrictive parser engine. This is why common tools like LIMS and ELNs exist for biotechnology. Having worked for vendors in this space for well over a decade, I've seen firsthand how much effort and treasure clients put toward the simple task of nomenclature harmonisation, for fear of this very problem. Months are spent negotiating between teams on h...
Read more on page 20 →