Feynman could have saved your life

 We had Covid vaccines since March 2020, and yet we sat around all year, letting the pandemic develop into a major catastrophe, waiting for nothing but the permission of evil bureaucrats to use it.

Well, not quite. There is a huge difference between discovering the recipe for something in the lab and producing enough of it to protect billions of people. As we are seeing, nearly a year later the limited production of the vaccine is still a major bottleneck to public-scale inoculation. 

Some idiot on Twitter seemed to think that the problem was evil drug companies hoarding their recipes and this could all be solved by having everybody join in and produce vaccines. As often happens, the resulting Twitstorm provoked someone who knew something about it to speak up. In this case, it was Derek Lowe, pharma industry expert and blogger for Science magazine. Here is his post Myths of Vaccine Manufacturing.

To sum up the key point of the blog, Lowe lists 5 major step in producing a vaccine of the new mRNA type:

1. Produce the appropriate stretch of DNA
2. Produce mRNA from your DNA 
3. Produce the lipids that you need for the formulation. 
4. Take your mRNA and your lipids and combine these into lipid nanoparticles (LNPs). 
5. Combine the LNPs with the other components and fill vials.
6. Get vials into trays, packages, boxes, trucks, etc.

To make a long story short (do read the original and the mostly informative comments if you are interested), the major bottleneck is step 4. As Lowe puts it, "Everyone is almost certainly having to use some sort of specially-built microfluidics device to get this to happen .... Microfluidics (a hot area of research for some years now) involves liquid flow through very small channels, allowing for precise mixing and timing on a very small scale.... My own guess as to what such a Vaccine Machine involves is a large number of very small reaction chambers, running in parallel, that have equally small and very precisely controlled flows of the mRNA and the various lipid components heading into them. You will have to control the flow rates, the concentrations, the temperature, and who knows what else, and you can be sure that the channel sizes and the size and shape of the mixing chambers are critical as well. These will be special-purpose bespoke machines, and if you ask other drug companies if they have one sitting around, the answer will be “Of course not”. This is not anything close to a traditional drug manufacturing process."

So I went and looked at what these microfluidics machines for producing lipid nanoparticles (LNPs) looked like. What's the scale, how close to nanotech do you have to be, etc, etc.

from ACS Omega 2018, 3, 5, 5044–5051

So here's a sketch of the kind of thing we are talking about, and a scale of the nanoparticles themselves, which tend to range from 20 to 100 nanometers in size.

Here's a closer look at the actual gadget:

Although it produces ~50 nm LNPs, the gadget itself is micro-scale, not nanoscale, technology. They are generally made using photolithography on the same kind of machines as computer chips. This is a tricky and complex process, with a lot of planning, work, and development between even a well-specified design and a usable product.

So what if we had a technology that could produce things like this right off the shop floor? If the scale were millimeters instead of microns, you could make one of these in an hour in your garage with a slab of MDF and a router. I could print one tenth of that scale on my 3D printer, and there are printers out there that could beat that by another factor of 10.

If we had full-fledged nanotech now, none of this would matter; after all a LNP is just an arrangement of atoms. But what struck me when I read about this bottleneck was forcibly to be reminded of this passage in Where is my Flying Car, Chapter 14:

Is it Worth Starting Now?

Surely, you will say, it would have been wonderful if back in 1960 people had taken Feynman seriously and really tried the Feynman path: we’d have the full-fledged paraphernalia of real, live molecular machinery now, with everything ranging from countertop replicators to cell-repair machines.

After all, it’s been 55 years. The 10 factor-of-4 scale reductions to make up the factor-of-a-million scale reduction from a meter-scale system with centimeter parts to a micron-scale system with 10-nanometer parts, could have been done at a leisurely 5 years per step—plenty of time to improve tolerances, do experiments, invent new techniques.

But now it’s too late. We have major investment and experimentation and development in nanotech of the bottom-up form. We have Drexler’s PNAS paper telling us that the way to molecular manufacturing is by protein design. We have a wide variety of new techniques with scanning probes to read and modify surfaces at the atomic level. We have DNA origami producing arbitrary patterns and even some 3-D shapes. We even have protein engineering producing the beginnings of usable objects, such as frames and boxes.

Surely by the time a Feynman Path, started now, could get to molecular scale, the existing efforts, including the pathways described in the Roadmap for Productive Nanosystems, would have succeeded, leaving us with a relatively useless millimeter-sized system with 10-micron sized parts?

No—as the old serials would put it, a thousand times no.

To begin with, a millimeter-sized system with 10-micron sized parts is far from useless. Imagine current-day MEMS but with the catalog and capabilities of a full machine shop, bearings that worked, sliders, powerful motors, robot arms and hands. The medical applications alone would be staggering. ...

That's where we should have been, at the very least. But of course we would still have to wait for permission.