The Computer Scientist's Guide to Designing mRNA Vaccines

Part 3: Sequence Optimization

3.3. Folding & Minimum Free Energy

As previously mentioned, the main contribution of Zuker & Stiegler (1981) was that different structures have different contributions to the minimum free energy of a folding configuration. The Nearest Neighbor Database contains records of these energy parameters, with the most recent published results from the Turner group in 2004.

The core idea behind current folding methods is that the structures contribute the same values to the free energy of a configuration, regardless of what distant parts are present. For example, base pairs contribute to free energy as Watson-Crick-Franklin Helices, where they have a ‘stacking’ effect based on their neighbors. Still, their energy contribution is not affected by a base pair 100 bases down the line. The parameters also impose penalties on certain base pairs or dangling ends according to rules determined experimentally.

Hairpin loops are another very common structure. They form when a series of base pairs ‘ends’ with a loop formed by unbonded bases. The loops can have as few as 3 bases, but can grow much bigger. Other loops include the multibranch one (which works as a junction) and bulge loops, which form a ‘bulge’ on a helix.

Overall, the computed minimum free energy is related to a high degree to the measured minimum free energy, which directly determines an mRNA molecule’s stability, and, as such, half-life. Vaccine designers should aim to increase a molecule’s stability, as it leads to longer shelf life. mRNA molecules that remain in a cell for longer are translated more, leading to greater protein output from the same amount of mRNA.