AlphaFold Didn’t Kill MX but 4th-Gen Synchrotrons Will

Every time a synchrotron moves from 3rd to 4th gen, the number of relevant MX beamlines users drops, the whole thing is on life support.

MX beamline were not out done by AlphaFold because they still had inertia. But the real killer will not come from AI, it comes from the ring upgrades. When a synchrotron shuts down for an upgrade, people say one year but it’s actually about two. In that time, MX users disappear, they ship crystals somewhere else and figure out how to work remotely. When the light comes back on, many never return. What replaces them isn’t a new generation. It is no generation.

In the U.S, the human capital for MX is drying up, there’s no cohort of 25-year-olds begging for 2 a.m. beamtime slots. The few that remain drift into analysis, modeling, or commercial R&D. MX in America is mostly carried by an older generation with habits that predate modern detectors. How many of the old guard still collect data at 1 degree per sec? At 4th gen synchrotron thats like asking a supersonic jet to taxi across the Atlantic. And yet the beamlines keep getting better: I bet the last horse-drawn carriage was a real beauty too.

So what is MX now? It’s confirmation. Fold-checking. We helped build AlphaFold’s training data (it seems so unfair) Room temperature dynamic, that’s real, that could carry forward. But it’s not what most beamlines are set up to do, and the US community that might do it isn’t really there yet.

Fragment screening is the diamond in the experiment rough, pharma wants answers. XChem thrives, creating the training data that will be its demise as well. But another wave is coming for them too. Faster tools, better proxies, AI-led design. When that hits, “screening” will need a new defense, confirmation machines like the rest of MX probably

The 4th generation of beamlines will still be here. The question is which science deserves the space, and should it be MX.