This blog is no longer active, you can find my new stuff at:,,, and

The Limits Of Medicine - Part 1 - Small Molecules

Much of medicine relies on what is dubbed "small molecules".

"Molecules" since they are atoms tightly bound together, forming what is easily seen as a unitary whole, as opposed to e.g. lipoproteins, which are synergistic ensembles held together by much weaker forces (pun not intended).

"Small", in that they are, well... lightweight. But for all intents and purposes, size and weight don't separate quite the same way at the molecular level. So whatever.

This class includes drugs such as all NSAIDs, all antibiotics, all antihistamines, almost every single drug a psychologist or even a psychiatrist is allowed to prescribe, virtually all supplements, sleep aids, wakefulness aids, every single schedule I through III drugs (i.e. fun and insightful drugs), almost all anesthetics, and almost all anti-parasitic drugs, etc.

Unless you are old or chronically sick, it's likely that the only drug you've ever taken that hasn't been a small molecule is a vaccine. Indeed, drugs that aren't small molecules are so strange and rare we usually don't think of them as "drugs" but as a separate entity: vaccines, monoclonal antibodies, anabolic steroids.

But, like, the vast majority of molecules found in our body, and in all of organic life, are not classified as "small". And the vast majority of the things doing something interesting are not really molecules, but more so fuzzy complexes of molecules (ribosomes, lysosomes, lipoproteins, membranes).

So why are virtually all drugs small molecules? Prima facei we'd expect most of them to be complexes made up of dozens to thousands of very large molecules.

The answer lies in several things:

Let's look at each of these aspects. None are unique to small molecules and not all small molecules bear all of them, but they are all traits significantly more likely to be found in small molecules.

Easy to produce

Producing a protein is hard, you have to get a gene sequence for the protein you want to produce, create some genetically modified organism (usually yeast) with a zillion copies of that genes, let it breed, extract the protein, makes super-duper-sure all potentially dangerous compounds are separated.

Along this process, you will have issues at every step, from errors in creating the DNA, to errors in creating the proteins to potential "errors" (mainly due to environmental contaminants) in how the protein folds and what exactly it contains, to errors at separation.

All of this is much harder and involves much more trial and error than simple compounds (e.g. most small molecules); For which we have some vague resemblance of "laws" in the form of classical and biochemistry. When it comes to proteins they are complex enough that predicting their behavior is often an inconclusive matter.

Small molecules, on the other hand, are often found in relevant quantities ready-made in organisms that are cheap to propagate (e.g. garden plants) and can be extracted using the most brutal of methods.

You can get a good approximation of thousands of life-saving drugs by the simple process of:

  1. Break down a plant with your bare hands in tiny pieces (ideally use pestle and mortar)

  2. Throw it in a bottle with gasoline and shake.

  3. Put pipe cleaner (H₂SO₄) in another bottle and throw in some table salt (NaCl)

  4. Connect the two with a tube, wait, pour in a pan, wait some more for gasoline to evaporate

Note: Don't try this at home, hydrochloric acid gas will melt your face and gasoline can explode, just buy your drugs from a reputable dealer (or pharmacist/doctor if you are really desperate and need a fix).

Ok, granted, most extractions are much more complex than this, but still, we've gone from vague hand waving with complex concepts like "building DNA" and "genetically modifying yeast" and jumping over 100 steps each of which took 10 PhDs to design to "here's a step by step 20 seconds guide to doing it in an ill-equipped kitchen".

In practice, for most small molecule drugs, the reactions producing them might be simple enough that we can skip the extraction step entirely and just synthesize them ourselves instead of extracting them. The synthesis of many arbitrary compositions for common large molecules (DNA, RNA and many types of proteins) has also become possible, but this is a rather common advancement, and humans of the 20th century would have been flabbergasted by the idea of an RNA printer.

Simple to administer

There are 2 major hurdles a drug must overcome in order to take effect:

But, usually, there are also two extra bonus ones that are super nice:

If you eat any complex protein (e.g. the kind that you put in a vaccine) it will be broken down into component amino acids in the stomach and small intestine. If you protect it with a capsule it won't be absorbed. If you design a fancy lipid capsule to facilitate absorption (or give it IV) it will be savagely attacked by white blood cells, immediately captured and digested in cell lysosomes, and pulverized down to glucose by the liver as a top priority. And we've not even gotten into whether it crosses a run of the mill cellular membrane (not that hard) or whether it crosses the blood-brain barrier (extremely unlikely for anything with more than triple digits atoms in its composition)

If you want this to be RNA or DNA now you've got the extra challenge of crossing a nuclear membrane ... there are a few ways to do it, but for all but a few niches they are as complex as "design a deactivated virus to carry it".

On the other hand, you can basically stare the wrong way at small molecule salt and it will perfuse itself into every living ounce of tissue in your body.

How can you administer aspirin? Swallow it_. Need a capsule? Not really_. Another way? inject it_. Do I have to hit a blood vessel? Naaah_. What if I want to look cool? _Ground it up and snort it. Can I snort it off someone's naked body at a party? _Yeah, but be quick, it also gets absorbed through the skin. Which tissues does it get into? _All of them. What if I'm transported back 5000 years in the past? _Just boil some willow root and drink it, you'll be fine.

Cheap to store

Small molecules are often content with just "existing" for a very long time, provided room temperature and lack of light, solvents, or water they can last mainly unaltered for a lifetime.

Most proteins, viruses, and lipo-whatever complexes used in medicine, on the other hand, require anything from being stored in a fridge and used within 3-6 months, to being stored at -120 degrees celsius and used within days.

This is not always the case, lysergic acid is a tiny yet infamously unstable substance and ApoB is a gigantic and infamously stable protein. But as a rule of thumb, the smaller the size and the higher the mass, the more stable a molecule is.

Homogenous in effect

Proteins are complex, everyone's are a bit different and their interactions produce different complexes and different epigenomes and those lead to yet more different levels of all proteins and it's all very loopy and head-scratchy.

Metals are simple, you've got like 11 of them and they all do basically the same thing in everyone, and the levels vary by a factor of like 2 or 5 between individuals, but not 1000 or 10^9 or +/- infinity. If you've got too much of them, the kidney usually filters them out without much issue, at most you end up losing some water. They are well preserved and well-tolerated in circulation.

Coincidentally, metals are small molecules and proteins aren't.

A life-saving protein can become lethal if you up the dosage to just 2-3x the levels, given previously mentioned large individual difference this essentially means that a lot of large-molecules would have to be administered by taking effects into account, continuously monitoring on the scale of seconds or minutes.

Currently, we get around this by not using proteins that are too risky and making use of fuzzy measures. People's tolerance to hGH is pretty flat, but it can still vary by a factor of dozens. The workaround for this is to start slow, tell people to increase the dosage, and stop when they feel iffy from it. But "feel iffy" is not a very quantitative endpoint, and relies a lot on individual interpretation, which is famously unreliable.

But to start using the interesting stuff we'd need a much tighter feedback loop, something closer to a device attached to a small catheter analyzing blood samples for dozens of markers every few [mili]seconds, a second catheter inserting a small amount of the drug and an algorithm regulating the dose based on the response in real-time.

All of our experience running clinical trials would be moot as well. Since a lot of people might just not respond well to these. Current best practices indicate being careful against selecting sub-groups that respond well. But to find the effect in protein-based medicine the most likely approach would be to find sub-groups that had an excellent response. You can still control for chance in this situation, but it gets much more complex

Quick to act

Finally, small molecules usually act pretty quickly.

They have a target, they bind to it, the target reacts, effects happen.

Small molecules usually modulate an existing process, often by binding to a receptor or enzyme, many times in an inhibitory fashion. But it's very hard for them to have a constructive or "additive" effect.

This is, in principle, a good thing, It makes using them in situations where they do work fairly easy.

Things like viral vectors inserting DNA plasmids into your cells... that takes a while to manifest, especially since there's no way to "turn it off" after 10 seconds, it'll be there for a few weeks, so you need to take into account the amount produced over time.

Wanna determine side effects before a long course? Though luck, inject a smaller quantity and wait a few weeks.

This might appear in contrast with what I just said previously, but it isn't, therein lies the problem. Often enough the marker you are monitoring to assess tolerance and dosage is not the final treatment outcome.

So you are stuck in a situation where for most large molecules you have to use a complex process to determine the dosage, then often enough wait weeks or months to see the desired effect, then rinse and repeat if it didn't work. Whereas with small molecules you just give the maximum safe dose and wait a few hours or days, and if it didn't work you give a slightly larger slightly unsafe dose or change meds and try again.

Of course, this is because the targets of large molecules are often more sophisticated and can't be reached with small ones. You can't restore genetic or immune integrity with simple substances, but you can with gene therapy.

Why is this a limitation?

Not only are small molecules limited, but thinking in terms of them limits the mindset of doctors, researchers and pharmaceutical companies.

Starting with prescription, where the dynamics of proteins, peptide hormones, and enzymes make them much more tricky. They can't just be prescribed on a "take 3 of these a day at mealtimes" basis, but rather, they require careful monitoring of the feedback loops they partake in.

Indeed, modern medicine has "burnt" many patients by prescribing the peptide hormone insulin in quite literally the "3 times a day with meals" fashion. Whereas the proper mechanism of administration would have been in tandem with monitoring of glucose. This lead to many becoming irreversibly "addicted" to the substance, no longer capable of producing or regaining the ability to produce it endogenously.

I use the past tense though, recently CGMs (continuous glucose monitors) are being used to administer insulin only when needed and in the quantities required, together with protocols for fasting and ketogenic eating to regain insulin sensitivity. I'm by no means in the vanguard here figuring this out.

The prescription of complex drugs will be aperiodic and heavily depend on the patient's demographic and monitoring. This is why, for example, boosters for sars-cov-2 and hepatitis B are always being tweaked based on the most active virus strains and administered only when results on various antibody assays indicate suboptimal leve... Oh? What? Really? Like, everyone? The exact same sequence used 40 years ago!? Really... ? Ok, so maybe there's still a bit of a learning curve.

But there's also the step of procurement. Whereas small molecules are best produced in a "centralized" fashion, this is not always the case with complex drugs, which might have a short shelf life and come with the problem of production variability no matter what's tried.

Might this mean that we should aim for local labs producing heavily unstable compounds on a need-to-administer basis? Or maybe a same-day production-supply mechanism for certain substances? Hard to tell, but the problem is currently not even being considered.

Finally, there's the issue of deciding who needs to take what. In tandem with continuous monitoring and "custom" orders, comes the possibility of designing "large molecules" for every individual patient, with slight variation specific to their DNA and bloodwork.

Would this require in-vitro experiments using patient tissue? Machine learning algorithms to design the drugs? Biokinetic models to test fundamental interactions? A new decision about the preferred administration method for every individual substance and patient based on their unique traits?

But it's not all doom and gloom

As I said in my Insulin example, medicine is slowly undergoing the process of learning how to work with complex molecules.

In parallel with normal medicine, I'd like to think that biohackers with the ability to custom order whatever they can dream of, will lead the way with self-experimentation. Figuring out the limits and benefits from individualized design and continuous monitoring.

For now, this is mainly click-bait, people injecting a bioluminescence gene with CRISPR kinda stuff, but some of it isn't. As an example, the guy from ThoughtEmporium self-designed a therapy to get rid of lactose intolerance (though the solution is not permanent, it lasts for a few months). Hundreds of other such people are engaging in similar experiments, and the more people do it, the more resources become available, the easier it will get, and the better the ROI.

As this happens, social acceptance will follow and pharmaceutical companies will get more on the deal.

It's not just lone loonies. There are entire clinics (with stunning results) dedicated to stem cell therapy, which create custom preparations using patients' own tissue and inject them at the specific injury sites with a case-by-case substrate composition to encourage healing. Not to mention techniques like PRP, a prime example of medicine using complex substances (cells) which is now available at every street corner and used for everything from hair loss to gum pain. Arguably, these people are still "the loonies" by mainstream standard, but they are certainly getting fairly close to widespread acceptance.

Also, we shouldn't forget that, while small molecules are limited, we are far from depleting their usage. Out of every single bioactive molecule that you could create, it's likely that just 0.0x% were ever tested in an animal, and out of those only 0.0x% were ever tested in a human, and out of those only 0.0x% underwent the rigorous trials needed to become an approved drug. The same techniques that will help us employ more complex molecules and constructs could well lead to a revolution in the way we administer and find small molecules.

Part 2 →

Published on: 2022-12-01



twitter logo
Share this article on twitter
 linkedin logo
Share this article on linkedin
Fb logo
Share this article on facebook