How AI is helping researchers develop antibiotics to fight drug-resistant infections

Amna Nawaz:

Drug-resistant infections are a major public health threat around the world, responsible for more than a million deaths each year. So scientists are constantly trying to find and develop new antibiotics.

Miles O'Brien reports on how researchers now say A.I. is helping to speed up their search.

Miles O'Brien:

This is the front line in a biological arms race to salvage the crumbling foundation of modern medicine, antibiotics. They make surgery routine, protect cancer patients, and turn once deadly infections into minor inconveniences.

Man:

Let's take a look at gonorrhea first.

Miles O'Brien:

The discovery of penicillin changed everything, not least the treatment of sexually transmitted disease.

Man:

It is a great boon to the private physician, the clinic, and, of course, to the patient.

Miles O'Brien:

But success comes with a fatal paradox.

Dr. Melis Anahtar, Massachusetts General Hospital:

The more we deploy this lifesaving medicine, the less effective they are in the long term.

Miles O'Brien:

Melis Anahtar is a clinical microbiologist at Massachusetts General Hospital.

Dr. Melis Anahtar:

It's unlike any other drug where, when we use antibiotics, we, by definition, lose them because we're in this constant race with bacteria, where the bacteria can evolve resistance to our antibiotics in real time.

Miles O'Brien:

Bacteria develop resistance to antibiotics through a simple process of evolution. In any infection, there are millions of bacteria, and some have mutations that help them survive a drug.

When antibiotics are used, they kill the vulnerable bacteria, but the resistant ones survive, multiply, and spread. Over time, these resistant strains become dominant, making the drugs less effective or even useless.

There's kind of this never-ending war.

Dr. Melis Anahtar:

Absolutely.

Miles O'Brien:

These microorganisms are -- they're just not going to quit.

Dr. Melis Anahtar:

They do not quit.

Miles O'Brien:

But neither does she or her colleagues here at the Broad Institute of MIT and Harvard.

Historically, to find and test new antibiotics, researchers would gather up some molecules stored in a deep frozen library of compounds and then apply them one by one to a pathogen to see what can either stunt its growth or kill it outright.

Biomedical engineer Jim Collins runs the lab.

Jim Collins:

It's a laborious process. It's very much kind of searching for a needle in a haystack that's an expensive haystack.

Miles O'Brien:

They would find a promising molecule less than 1 percent of the time. Enter artificial intelligence. Collins and his team trained a deep neural network to analyze the chemical structures of molecules.

Jim Collins:

And bond by bond, substructure by substructure, so kind of these ball sticks that we all remember from our high school chemistry days, could associate those properties with being antibacterial or not antibacterial.

And now a model was trained that you could feed a new compound structure, these balls and sticks, and the model could calculate, could this make for a good antibiotic?

Miles O'Brien:

They applied the A.I. to the library of 6,000 compounds here at Broad to find molecules that would make for antibiotics that are effective, are not toxic to human cells, and have not yet been discovered.

Jim Collins:

Of the 6,000, only one molecule satisfied all three criteria, and it's a molecule we call halicin. And halicin turns out to be a remarkably potent new antibiotic that kills multidrug-resistant, extensively drug-resistant and pan-resistant bacteria through a new mechanism of action.

Miles O'Brien:

Then they deployed A.I. and computing power to virtually generate and screen 70 billion theoretical molecules to test how they might behave.

In this case, A.I. is doing something computational chemist Andreas Luttens does intuitively, instantly seeing what molecules might work against a pathogen based on how the balls and sticks line up.

Andreas Luttens, Computational Chemist:

I usually work with a bucket system and -- or a scoring system. Like, there's stuff that I really like, stuff that I hate. No, no, no, no, no, maybe, no.

Miles O'Brien:

This is like small molecule Tinder. You kind of swipe right, right?

Andreas Luttens:

Yes. Yes, exactly, yes.

Miles O'Brien:

Yes.

Andreas Luttens:

It is a molecular dating app. You're trying to find compounds that you like, and you can go quite fast in the selection procedure.

(Laughter)

Miles O'Brien:

He may be fast, but he can't match the scale and persistence of a machine.

Andreas Luttens:

Depending on how much coffee I get, how long I can stay awake, but it will beat me eventually.

Miles O'Brien:

Which brings us back to Melis Anahtar and the pathogen she is focused on, Neisseria gonorrhoeae.

That's gonorrhoeae right there, huh?

Dr. Melis Anahtar:

This is gonorrhoeae.

Miles O'Brien:

Wow. Looks nasty.

Dr. Melis Anahtar:

So you see these little colonies, grayish.

Miles O'Brien:

Yes.

Untreated, the sexually transmitted disease can escalate into serious, sometimes irreversible health problems. The bacterium is resistant to nearly everything, outsmarting new drugs about every five years or so.

The currently prescribed antibiotic Ceftriaxone is nearing the end of its efficacy. Finding something new and potent is an urgent problem. The A.I. system screened 45 million chemical fragments from that vast universe. The most promising chemical seed was used to generate an additional seven million candidates.

After rigorous filtering, two of these new compounds were synthesized and tested against real bacteria in the lab.

Dr. Melis Anahtar:

Is it really killing the bug in vitro? And is it not harming human cells? Pink means there's bacterial growth, and blue means that the growth is inhibited. So we want to see lots of blue.

Miles O'Brien:

This looks like a home run here.

Dr. Melis Anahtar:

So this one looks good. This one was not as successful.

Miles O'Brien:

Oh, not so good, yes.

In the end, there was one novel compound that killed drug-resistant gonorrhea without causing serious harm to human cells.

Dr. Melis Anahtar:

Not only can we find these antibacterial compounds, but they're actually inhibiting new targets. This was created from scratch based on what it learned from existing small molecules and drugs.

Miles O'Brien:

Globally, drug-resistant infections kill more than one million people each year. And if nothing changes, the experts predict that number will increase by 50 percent by 2050.

So is resistance right now moving faster than the research to try to address it?

Jim Collins:

Resistance had been developing faster than the research and development that had been under way. But I believe that this infusion of A.I. has now changed the game. We now have tools that have dramatically expanded our ability to both discover and design new antibiotics.

Miles O'Brien:

None of this will increase the speed of clinical trials in human beings, nor should it. And it does nothing to incentivize big pharma to manufacture new antibiotics, which don't generate big profits.

But artificial intelligence might be one way to begin recharging a crucial pipeline that has dried up.

For the "PBS News Hour," I'm Miles O'Brien in Cambridge, Massachusetts.