I’ve been reading Gregory Zuckerman’s A Shot to Save the World, on the science and history of the COVID-19 vaccine. This passage on Jon Wolff stood out to me:

Wolff wondered: What if we could synthesize and deliver normal DNA or mRNA, without the mistakes, straight into human cells to replace defective genes in his patients? Maybe this way, functional proteins could be created.

When Wolff shared his simple-yet-outlandish idea with peers he didn’t exactly blow them away with his originality. Of course it had occurred to them that injecting normal DNA or mRNA might be a way to help those suffering from various diseases; the question was how to do it and whether it was even possible. Scientists were beginning to appreciate that if DNA or mRNA could be delivered into the body’s cells, they could theoretically read the instructions and make the corresponding protein, potentially curing genetic disease. It was all part of a new potential method of treatment being described as gene therapy.

But hardly anyone was considering injecting straight, or “naked,” DNA, an exercise that likely was useless, if not dangerous. There’s a reason it’s a two-step process—DNA to mRNA to proteins. Within a cell, DNA has to be transcribed into mRNA before it can create useful proteins. It’s not easy getting DNA into a cell to begin the protein-creation process. Since DNA is a huge, negatively charged molecule, most researchers expected the cell’s membrane to block it from entering the cell. Besides, messing around with DNA, which is the body’s permanent instructions, seemed risky.

Wolff’s colleagues were even more certain that injecting messenger RNA wouldn’t work, largely because it’s such an unstable molecule. After moving within the cell to the cytoplasm and providing necessary instructions to create proteins, mRNA is usually degraded, broken down, in a matter of hours. Just as a spare tire can hold up on a short drive but will break down on an arduous cross-country trip, researchers agreed it was folly to expect mRNA to survive a solo trip into the cell, in the hope that it could produce sufficient proteins. Everyone knew that the moment mRNA was injected, it would come into contact with bodily fluids chock-full of enzymes that would immediately chop it up.

Wolff understood the resistance to his idea, but he didn’t want to give up. He began looking for ways to get DNA and mRNA into cells to facilitate gene therapy. He became consumed by the intellectual challenge, unable to let it go.

“This was something he never stopped thinking about . . . the scientific puzzle of getting genes into cells,” Katalin says.

Recent advances had shown that by mixing DNA or mRNA with certain liquids or fatty lipids prior to adding them to cells, it was possible to induce the cells to absorb the DNA or mRNA. The lipid packaging seemed to protect these two nucleic acids, helping to carry the molecules and their genetic messages through the cell’s membrane. But most of the research had been done only on cells growing in a Petri dish. Wolff was an experimentalist, though, so he thought he’d go ahead and inject mice. He didn’t have much to lose. It helped that he had ample funding from various sources, including a lawyer for James Watson—the molecular biologist who helped identify the structure of the DNA molecule—who had a son with muscular dystrophy.

Wolff set up a series of experiments to test whether DNA or mRNA injections packaged in crude lipids could create proteins in mice, comparing those shots with a control group of straight DNA and mRNA shots without the lipids.

What Wolff saw was truly astounding. The shots of DNA and mRNA encased by lipids didn’t do much to the mice, a disappointing result that was in line with what the naysayers had predicted. But when Wolff directly injected DNA and mRNA into the leg muscles of mice, he found that they were actually successful in creating the desired proteins in the mice cells. The control group had somehow worked better than the experimental group.

In sciences like biology and machine learning where we understand little mechanistically, it pays to skip theorizing about promising ideas and just do the experiment. The corollary is that more smart people should work on making the feedback loops in these fields faster. With faster feedback loops, researchers don’t just get more done, they work on more important things.