The cost of doing science

I’m here to talk about why science is so expensive.  This was prompted by the recent news regarding Turing Pharmaceuticals, led by CEO Martin Shkreli.  Turing recently acquired exclusive rights to manufacture and sell pyrimethamine, an off-patent drug used primarily to treat parasitic protozoan infections.  In a move which Ayn Rand herself would probably describe as “heavy-handed,”  Shkreli has opted to increase the price of the drug over fifty-fold overnight*, somewhat disconcerting considering it’s used to keep AIDS patients from, you know, dying.

The issue garnered international attention, with many (many) calling this an example of the absurd pricing power granted to the pharmaceutical industry.  Except it’s really not a par for the course industry move, and just as many industry insiders are condemning the price hike.  I won’t delve into the details, as plenty of others much more knowledgeable than I already have.  I’m also not going to beat the “drug research is expensive” dead horse (spoiler: it is).  The point is, many people seem to cry foul**; “There’s no way R&D can be that expensive!”

Instead, I’m going to talk about why science in general is so expensive (second spoiler: it is) and where the associated costs come from.  Television seems to have popularized the idea of a genius scientific duo working in tandem and checking off a major scientific breakthrough (bioweapon vaccine, invisibility suit, quantum ion machine gun, etc.) in a montage-filled afternoon.  Unfortunately, that idea doesn’t mesh well with reality.

Science Cannot Exist in a Vacuum

Let’s start by examining the core of what science is, as an ideal.  What is science?  I’d argue that the answer to that question is quite simple: science is discovering new things and telling others about those discoveries.  That’s it.

All the impact factors, intellectual property, marketing, and publications?  Those are politics, or business; consequences of science, but decidedly not science.

What’s it take to do science?  Fundamentally, a scientist, of course.  You can’t research new materials without chemists, you can’t map the universe without astrophysicists, or the genome without geneticists.  But it turns out, you also can’t map the universe without computer scientists.  Likewise, you probably need a few synthetic chemists somewhere along the line to make probes for your geneticists.

Eighteenth century science could be done by a single bowtie-clad bearded guy with sufficiently deep pockets.  But modern science cannot.  Modern science is collaborative by nature.  To the scientists reading, this is a point I shouldn’t need to argue.  In the past week alone I have collaborated with: several materials scientists, a biologist, an inorganic chemist, a couple laser specialists, and at least one battery scientist.  And that’s not even an exhaustive list.

In order to properly do my job, I need to be surrounded by a dozen or so other scientists, who in turn, each need a dozen or so to support them.  My company employs in the ballpark of 100 scientists, and I’ve worked with almost all of them at one time or another.

And that’s where the baseline cost of doing science comes from.  To do  science, you need a network of scientists.  How big that network needs to be depends on how complicated and diverse the problems you are working on are.  And as much as we scientists generally love doing science, we don’t do it for free.

Scientists Need Tools

This seems obvious, but most non-scientists probably haven’t stopped to consider how much scientific equipment costs.  And I’m talking capital equipment, not stuff that gets used up like glassware and reagents.

Consider: I operate a Varian 300 MHz NMR spectrometer on a daily basis.  It’s an old instrument, but it’s in excellent working condition.  Despite being old, it still takes time and money to keep it up and running.  How much do you think it costs to operate for a year?  When you consider the service contract, preventative maintenance, cryogens, and time, I think it’s safe to say that it’s more than I make in a year.  And that’s a fixed cost; if I use it every single day, or if it sits in it’s air-conditioned room as a giant paperweight, that cost is the same.

Now imagine a larger company which may operate several such instruments.  They’d almost certainly have at least one full-time scientist charged with maintaining them, and that’s on top of the other costs.

And an NMR is actually a pretty simple instrument to take care of.  If you’ve got an LC/MS (a modern triple-quad will run around $150k), you now have tons of additional operating costs to consider, and you probably need one technician per 2-3 instruments just to keep them operational.  Same goes for GC/MS.

If you’re a materials scientist, you need an electron microscope.  Those cost a pretty penny.  Plus, they pull huge amounts of power to run.  If you’re doing any serious biology experimentation, you probably need a decent confocal microscope.  I have it on good authority that a single lens for one can set you back tens of thousands of dollars.

Power, service contracts, technicians, software licenses, instrumentation.  Everything adds up, real quick.  You could buy a townhouse for the cost of an 800 MHz NMR spectrometer.

In addition to the tools, you need a place to do science.  Fume hoods, bench space, ventilation.  You’ve got to keep the lights on and water flowing, and pay the building’s rent.  When you spend money on science, you’re spending money on this infrastructure, all of which needs to be paid for somehow.


Now you’ve got a sufficiently large group of interdependent scientists, all under the same roof, and all standing around with shiny new tools and instruments.  But they still need stuff to work with, and that means spending money on consumables.  Glassware, solvents, reagents, gases.  If you’re of the biological persuasion, cells and animal models.  And all of those are sold at a premium, because remember all that infrastructure I talked about earlier?  Yeah, the companies supplying researchers have all of that to deal with too, plus the added costs of manufacturing, quality assurance, regulatory compliance and shipping logistics.

All that means that a pair of scissors from Sigma Aldrich costs roughly fifteen times what it would cost at Staples.  But really, that’s just a somewhat egregious example that non-scientists can get their head around.  I’ve ordered my share of compounds which cost upwards of a thousand dollars for one milligram.  I’ll go through almost a 4-liter jug of acetone in a week, a liter or so of DCM, chloroform, or any of half a dozen other common solvents.  And if you’re running a lot of NMR (like I am) you’re going to chew through deuterated solvents, which cost around ten times what their isotopically abundant counterparts do.

Working with living organisms?  Well now everything, reagents included, needs to be sterile, which takes time and energy and is passed on to you in the form of higher prices.

Plus everything needs to be disposed of, and as you’re probably aware, you can’t just pour your reactions down the sink when you’ve finished with them.  Having chemical and biohazardous waste disposal companies on call to haul out the stuff you buy costs money too, sometimes even more than the cost of the stuff you originally bought (I’m looking at you, old lecture bottles).

Time, Time, and More Time

Those of you personally familiar with the scientific process will be intimately aware of the single largest cost of doing research: time.  Time is money, and science takes a lot of time.  Much, much more often than not, research leads to failure, which leads to slightly more informed failure, which, if you’re persistent and lucky might lead to success.  That’s not the fault of the researchers, that’s just the way it is.

And in order to do experiments, even ones that will invariably fail, you still need all those consumables talked about earlier.  I recently attempted to prepare a compound according to a slightly modified literature procedure that involved bubbling a fluorinated, gaseous compound through an anhydrous ammonia solution.  And guess what?  It didn’t work.  By all accounts it should have, but hey, research.  Over a thousand bucks for a cylinder of specialty gas, another couple hundred for the solvents and ammonia.  An entire day of careful planning and preparation.  And all of that gone in one shot on a failed*** experiment.  I had to order everything again, modify my procedure, and try again; the second time turned out to do the trick.

The point is: it cost twice as much in materials, took more than twice the amount of my time as originally planned, and set my progress back a week.  And that’s just a single experiment in the context of a much larger research program.  When you consider that even a modestly sized research program will have several scientists performing related but independent experiments, each of which carries a significant risk of failure, it’s easy to see how much time things can really take.

And that’s OK, because failure is a feature of the scientific process, not a bug.  Science cannot progress without failure.  Most of the time there’s really nothing that can be done about it, even under ideal conditions; sometimes things just don’t work.  But what many don’t realize is that failure incurs the same costs as success, and it takes a lot of failure to get to point where you can call a research program “successful.”

Ultimately this is why emergent technologies are so expensive.  This is why you see new drugs which cost close to $100,000 for a course of treatment.  It’s because for each Sovaldi, there were hundreds (or thousands) of candidate compounds which didn’t make the cut for one reason or another.  Every one of Tesla’s high-capacity lithium metal oxide batteries sold will be covering the cost of the undoubted thousands of experiments which led to their development.


*They’ve since back-peddled a bit on the exact price point, but not before very harsh and very public backlash

**To be clear, Shkreli and his ilk don’t seem to be doing any actual research with their profits

***There are no failed experiments, only new ways not to make the compound in question