How a key-shaped molecule picks the virus's lock express

Original: "Unraveling the Mechanism of Drug Binding to SARS-CoV-2 RNA Pseudoknot with Thermodynamics-Driven Machine Learning"
· Mariia Ivonina, Jakub Rydzewski
arXiv:2604.14906 · 2026-04-16 · CC BY 4.0 · 1 min
SARS-CoV-2 has a tricky RNA knot, and drugs act on it differently depending on how the knot is tied.

The RNA of SARS-CoV-2 is not just a chain but a complex structure. It contains a region that ties itself into a knot, much like a tangled shoelace. This knot helps the virus produce two proteins from a single stretch of genetic code. Without such a knot, the virus replicates less efficiently, so scientists are hunting for compounds that can undo it.

Computer simulations showed that three similar drugs act on the knot in different ways. If the knot is tied one way, the drugs break apart one part of it; if tied another way, they break two other parts. Moreover, the strength of this disruption matches exactly how well the drug stops the virus in a test tube. It's as if one key only fits one of two locks on the same door.

The acidity of the environment changes RNA behavior more than the type of drug. In an acidic environment, like inside a cell, the knot behaves differently than in a neutral one. Most previous calculations didn't account for this, and maybe that's why they were off.

Now scientists know: to design a drug against the virus, you must consider both the knot's shape and the acidity. It's like fixing a clock mechanism: it's not just about replacing the spring, but also knowing which position the gears are stuck in.

The knot in coronavirus RNA makes the ribosome (the cell's protein-building machine) sometimes slip one step backward: this glitch is what lets the virus get two proteins from one stretch of code.

In science fiction, viruses are often shown hacking into a cell like a hacker into a program. Here, scientists are literally looking for an 'antivirus' that prevents malicious code from executing.

Scientists
William Lawrence Bragg
Concepts
protein folding protein structure binding affinity Post-transcriptional regulation RNA self-supervised learning protein language model