Socialize Big Pharma Today. Save Your Life Tomorrow.

Deaths from antibiotic resistance will hit 10 million a year by 2050. But despite recent breakthroughs using artificial intelligence to discover antibiotics, the private sector doesn’t find it profitable enough to make new and better antibiotics. We’re all going to suffer unless the pharmaceutical sector is socialized.

A pharmacy technician grabs a bottle of drugs off a shelve at the central pharmacy of Intermountain Heathcare on September 10, 2018 in Midvale, Utah. George Frey / Getty

With coronavirus devastating people around the world, this might be a strange time for public health optimism, but it’s worth discussing the implications of a recent bit of good news. The world’s first antibiotic discovered by artificial intelligence, announced earlier this year, is genuinely a stunning breakthrough. It’s an example of the promise of machine learning finally delivering in a spectacular way.

The MIT researchers responsible for the new drug discovery approach named the antibiotic halicin named after HAL, the killer AI from 2001: A Space Odyssey. And halicin is a killer indeed, robustly effective against a great range of multidrug-resistant “superbug” strains of bacteria, including of Mycobacterium tuberculosis — responsible for tuberculosis, and two of the World Health Organization’s top three priority targets for pathogen research, Acinetobacter baumannii and Enterobacteriaceae, due to their resistance to carpabenems, a class of “last resort” antibiotics beyond which we have no defense left.

The researchers first trained an artificial neural network — a computational learning system that uses a series of simple but interconnected information-processing nodes that mimics the network of neurons that make up animal brains in order to identify relationships in a set of data — with a collection of a couple thousand molecules. The collection, included existing approved drugs but also other substances that we know work to disrupt bacterial activity. The model doesn’t have to be programmed with the expertise of molecular biologists who know how antibiotics work; instead, it learns patterns of which these human experts may not be aware.

Once this digital beagle was trained to sniff out these molecular patterns of “antibiotic-ness,” it was then tested by having to rifle through a library of some 6,000 molecules currently being investigated as possible candidates for treating different human diseases. They asked it to identify compounds that would be effective against E. coli, and picking out ones that had strong antibacterial activity but also that had a chemical structure unlike any current antibiotics.

Those identified were then in turn tested in the real world on mice and one was found not only to be effective against a broad range pathogens, but appeared to have low toxicity and — due to its novel mechanism of antibacterial activity — was much more robust against the development of antibiotic resistance.

Recent lab-based antibiotic discoveries have suffered from their relative similarity to existing antibacterial mechanisms. This means that evolutionary mutations need little more than a few days to bypass them, and antibiotic resistance quickly arises once again. With radically different antibacterial mechanisms identified by the algorithm, bacterial evolution is a bit more stumped. The researchers said that even thirty days later, they saw no signs of resistance to halicin.

Following this success, the research team then set the beagle off to hunt for more drugs and applied it to an even larger library of 1.5 billion compounds, initially on just over 100 million of them. Once again, some twenty-three candidates were identified, two of which appear especially effective.

The researchers now want to apply this technique to discover antibiotics that are more discriminating in the types of bacteria that they attack: killing the pathogens while leaving beneficial gut bacteria alone.

AIs have been applied to other aspects of antibiotic research before — what the researchers term “in silico” screening, as opposed to in vivo (studies conducted in living organisms) or in vitro (those conducted “in the glass” outside organisms via test tubes, petri dishes, flasks, etc). But previous models have still needed some bootstrapping through human assumptions; none have sufficiently accurate to identify brand new antibiotics without such assistance. James Collins, the bioengineer who led the team, reckons halicin is one of the most powerful antibiotics ever found.

In a paper in the journal Cell detailing their work, the researchers describe how they were driven to try out this method by “the decreasing development of new antibiotics in the private sector that has resulted from a lack of economic incentives.” If urgent action is not taken to both discover and develop new antibiotics, public health officials project that deaths from antibiotic resistance will hit 10 million a year by 2050.

This otherwise preventable annual calamity could arise all because it isn’t sufficiently profitable to research, test, and manufacture a commodity that if it works will only be purchased for a few weeks or months at most until the infection is gone, and works best the fewer people use it. Antibiotics work in the opposite direction to how the free market works.

Antibiotic discovery is also already just really, really difficult, and getting more so, with the same molecules being rediscovered over and over. New versions of existing antibiotics “results in substantially more failures than leads,” the authors of the halicin paper lament.

But fundamentally, the problem of antibiotic resistance comes from, as the researchers say: “the decreasing development of new antibiotics in the private sector that has resulted from a lack of economic incentives,” exacerbating the difficulty.

“We’re facing a growing crisis around antibiotic resistance, and this situation is being generated by both an increasing number of pathogens becoming resistant to existing antibiotics, and an anemic pipeline in the biotech and pharmaceutical industries for new antibiotics,” says James Collins, one of the paper’s authors.

Major pharmaceutical firms have abandoned all aspects of antibiotic development and production, because it makes no business sense to be producing a drug that works best the fewer the number of people that use it, and that is also only used for a few weeks or months at a time. Drugs for chronic diseases that have to be taken every day for the rest of a patient’s life are significantly more profitable. So Big Pharma largely left this space some three to four decades ago, preferring the greener pastures of more profitable therapeutics.

There are still small and medium-sized companies, often spun out of university or government labs, that have attempted to fill in the gap, but they tend not to have the capital or other resources needed to go beyond early research stages. Some researchers and some companies are focusing their efforts on alternatives to antibiotics such as bacteriophages, antibodies, probiotics, or lysins. These are all promising, but they too suffer from the same inherent market failure that antibiotics do.

So the researchers using machine-learning for antibiotic discovery are looking to this novel technique in order to boost the rate at which early leads for potential new antibiotics are identified and in so doing, radically reduce the cost. By slashing the capital requirements for early lead discovery, it is hoped that this could make antibiotic research, development, and manufacture more attractive to pharmaceutical companies.

A press release from MIT announcing the discovery suggests that this new technique has replaced that anemic pipeline with “a new pipeline,” perhaps even “a new paradigm” for drug discovery writ large. The researchers hope that the same approach could be used for other kinds of drugs that could be used against cancer or neurodegenerative diseases.

But this only ameliorates the problem. It doesn’t solve it.

We still run into the same inherent market failure that has been recounted now so many times, not just for antibacterial defense, but for anti-fungals, for vaccines, for neglected tropical diseases, for development of diagnostics appropriate to the developing world, for development of diagnostics for priority pathogens during non-outbreak periods.

The problem is that if a good or service is not profitable, or even insufficiently profitable, a company will not produce it. In fact, an officer of a company actually has a legal responsibility to not do so.

Three cheers for potentially making early lead discovery much cheaper with AI. But it still remains the case that regardless of how cheap part of the production process is, if a commodity isn’t sufficiently profitable, it still won’t be made.

And drug discovery is also only one part of the overall production cost. Clinical trials are incredibly expensive, and manufacture and distribution do not come for free either. The researchers are now hoping to partner with a pharmaceutical company to take halicin into clinical trials, or perhaps (as is perhaps more likely) a nonprofit organization.

We are increasingly seeing medical charities, intergovernmental nonprofit networks, and billionaire philanthropists filling in the growing gaps in the provision of global medical research, manufacture, and infrastructure. Report after report is issued recommending various kludges to this vast market failure, such as increased subsidy to private companies, expansion of funding for research grants, tax incentives, prize funds that link reward payments to the health impacts, accelerated approval procedures, advanced market commitments, or paying pharmaceutical companies a regular insurance premium to develop antibiotics when required.

In the last decade in response to the growing problem of antibiotic resistance, there have been more than fifty major new international and national initiatives trying to fix the market failure and incentivize antibiotic R&D and manufacture, from the Joint Programming Initiative on Antimicrobial Resistance (JPIAMR), to the New Drugs for Bad Bugs (ND4BB) program. The UN General Assembly has held high-level meetings and political leaders have increasingly spoken about antimicrobial resistance as a priority.

But none of these efforts, as welcome and helpful as most of them are, attack the root of the problem the way that decommodifying the pharmaceutical sector in its entirety does. In other words, the complete elimination of this sector from the market akin to how in many jurisdictions fire brigades are simply a public service. The analysts all take as given that we just need to find the right incentives for private companies. The possibility that the very existence of private companies in this sector are the problem is beyond the pale.

Some researchers and health economists have at least in the last few years dipped their toes into suggesting the creation of publicly owned pharma companies. But while this gets much closer to the core problem of decommodification, it still leaves the existing pharmaceutical firms standing.

We should not be afraid of the word nationalization. By government takeover of the sector as a whole, we can simply redirect monies made from profitable drugs to those that are not (and may never be), without worry of going out of business. This is the most efficient option. It is also the fairest. Why should we leave private companies to cherry-pick the profitable areas of drug development while the taxpayer gets stuck with the unprofitable areas?

But we cannot blame all the policy advisors, experts, researchers, and clinicians for avoiding what surely should be the most obvious and straightforward option. They studiously ignore the elephant in the room because, well, that’s crazy talk. Nationalization of the entire pharmaceutical sector? That’s, that’s, well, that’s socialism!

Yes. Yes, it is. And there is a parallel here to the conversation about democratic socialism within the debate over Medicare for All versus merely a public option. Thanks to years of a mass movement fighting for a single-payer health system in the United States, and the uncompromising insistence by Bernie Sanders that half measures simply do not produce the best health outcomes, are unfair, and more costly, we are now in the situation where almost everyone in the Democratic Party has to at least rhetorically concede that a public system of some description may be the best solution, but it just isn’t realistic yet.

What was dismissed until recently as inconceivable is now a serious option on the table, and all other figures are forced into an orbit around that policy framework. We need to do the same with pharma: make nationalization recognized as the default best option that everyone now has to respond to.

So, the lesson is that a similar scale of mass movement that has occurred for Medicare for All, but now for the nationalization of the pharmaceutical sector, must be built in the coming years. Of course, it must be recognized that the priority in the United States right now is Medicare for All, but spinning out of that once that battle is won, there needs to be a new movement around pharma as well.

Clinicians and researchers confronting antimicrobial resistance repeatedly describe the problem to be at least as great a threat as climate change, yet we do not see activists working on this issue with the same commitment or intensity. That needs to change.