C&EN’s Year in Pharma 2019

After a year in which deals were few and far between, major drug firms made up for lost time in 2019. This year, big pharma companies brought back the megamerger, inking several mammoth deals that will reconfigure the industry ranks. They also made a slew of midsize purchases meant to add technology or bolster a key therapeutic portfolio.

According to an analysis done by the investment firm SVB Leerink, global pharmaceutical merger and acquisition activity is on track to hit a 10-year peak in 2019. Leerink expects total transaction value for the year to reach or exceed $250 billion.

Bristol-Myers Squibb kicked off the bustling year by announcing it would pay $74 billion to acquire Celgene in a deal designed to create a leader in cancer and immunology, areas where the companies have complementary drug portfolios. That deal led to another sizable one: in August, Celgene agreed to sell its psoriasis treatment Otezla to Amgen for $13.4 billion, an arrangement intended to satisfy the US Federal Trade Commission’s concerns about the dominance of BMS and Celgene’s combined immunology franchise.

As the year unfolded, several other large deals were proposed. AbbVie, anticipating patent loss on its lucrative Humira immunology franchise, announced plans to buy the specialty pharmaceutical firm Allergan. And Pfizer unveiled plans to combine its Upjohn generics business with Mylan, forming a stand-alone firm with about $20 billion in annual revenues. Pfizer, meanwhile, beefed up its innovative medicine business with the $11.4 billion acquisition of Array BioPharma.

A few factors motivated this year’s surge in activity, says Glenn Hunzinger, pharma and life sciences deals leader at the consulting firm PwC. Among them was a shake-up in the leadership at many big pharma companies. Pfizer, Novartis, and Gilead Sciences, for example, all have new CEOs. “Once people get in their seat, they start to put their new strategy to work,” Hunzinger says.

And medium biotech firms were more affordable in 2019. In 2018, “the biotech markets were pretty rich, and a lot of deals at that time were a $10–$20 billion binary bet,” meaning early-stage drug candidates being acquired had an equal chance of succeeding or failing, Hunzinger says. But the market is less heady today, in part because the release of clinical data for hyped drug candidates is bringing values down to reality. Those same assets this year cost closer to $5 billion or $10 billion. “That’s a much more palatable risk to take,” Hunzinger says.

The analyst expects 2020 to continue to be an active year for M&A. One or two more megamergers are possible, he says, and companies will continue to build their pipelines through midsized purchases.

Credit: Brendan Mcdermid/Reuters/Newscom (Gottlieb); Ron Sachs/Cnp/Admedia/Sipa/Newscom (Sharpless); Bill Clark/CQ Roll Call/Newscom (Hahn)

Stephen Hahn (right) is expected to take over the leadership of the US Food and Drug Administration. Previous leaders include acting commissioner Norman Sharpless (center) and Scott Gottlieb (left).

Vaping deaths. Drug shortages. Contaminated medicines. Opioids.

These are some of the biggest issues that the US Food and Drug Administration has dealt with in 2019, a year that saw one commissioner resign, one acting commissioner reassigned, and a new head nominated.

Stephen M. Hahn is expected to be confirmed as the FDA’s new commissioner in the next few weeks. In addition to these public health issues, he will have to deal with a pharmaceutical and biotech industry that wants fewer barriers to getting drugs to market and a public fed up with high drug prices. Hahn would be the fifth person to lead the agency since President Donald J. Trump took office in 2017.

Hahn would replace Scott Gottlieb, who took office in May 2017 and resigned roughly 2 years later. Gottlieb was known for his strong public and personal stances on issues like smoking and opioids. He sang the praises of e-cigarettes as a way to prevent cigarette-related illnesses but later backtracked on his support as large numbers of teens and young adults started vaping and getting sick. Gottlieb proposed reexamining the efficacy of opioid painkillers as a first step toward potentially changing the way they are regulated and even pulling some off shelves.

Between Gottlieb and Hahn was Norman E. “Ned” Sharpless, who spent approximately 7 months as acting head while Trump considered a replacement. Sharpless followed suit on Gottlieb’s efforts to regulate the vaping industry, proposing guidance that would end flavored e-cigarettes and suggesting that several chemicals found in the devices be added to a list of harmful chemicals in tobacco products. Brett Giroir served briefly as interim commissioner after Hahn was nominated and Sharpless was reassigned to the National Cancer Institute.

It’s not clear if Hahn, who Trump nominated at the last possible legal minute, will also follow in Gottlieb’s footsteps or forge an entirely different path.

His nomination was generally met with accolades.

“He seems to understand what his responsibility is, what his challenge is, and I think people are willing to give him the benefit of the doubt,” says Bernard Munos, who spoke to C&EN before the confirmation hearings. Munos is a senior fellow at FasterCures, a think tank trying to bring drugs to market more quickly.

The revolving door at the top of the agency might look troublesome, Munos says, but the agency’s day-to-day work will likely not change, at least as it relates to drug approvals.

“It used to be seen, back 15 years ago, as an impediment to innovation. Today, it’s widely recognized as a catalyst for innovation,” Munos says about the FDA.

Hahn is a radiation oncologist and in that role has been part of dozens of studies testing the efficacy of various cancer treatments. He also has leadership and crisis experience: he steered the University of Texas MD Anderson Cancer Center out of financial and legal trouble, and while he was at the University of Pennsylvania, he took responsibility for shoddy treatment of veterans with cancer.

The commissioner’s office has seen a lot of turnover since Trump took office. Depending on Trump’s reelection success, Hahn may not have a lot of time to realize his agenda, no matter how ambitious it might be.

Credit: Amgen

The closely watched cancer drug candidate AMG 510 nestled into a pocket on KRAS G12C

In June, the drug industry held its breath as a cancer clinician took the stage at a major oncology conference and unveiled the first results on AMG 510, a small molecule that blocks a mutant form of KRAS. That data set represented the culmination of 30 years spent chasing KRAS, which is the most commonly mutated gene in cancer that until now has been impossible to drug.

AMG 510, discovered by chemists at Amgen, is one of four compounds in the clinic that directly inhibit KRAS G12C, a mutant form of the protein found in some 13% of lung cancers and 3% of colon cancers. Mirati Therapeutics, Eli Lilly and Company, and Johnson & Johnson have also begun human studies of KRAS G12C inhibitors.

Throughout the year, Amgen and Mirati offered early looks at how their drugs work in a few dozen patients with various tumors. Oncologists, drug developers, and industry watchers alike carefully analyze each report to try to understand the activity of this new class of compounds. Preliminary data unveiled at the June conference and elsewhere suggest that KRAS G12C inhibitors cause tumors to shrink in people with lung cancer but on their own might not be enough to overcome colon cancer.

The biggest open questions are how long the early effects will last and whether the benefits of the drugs can be extended if they are combined with other cancer therapies. KRAS is a central node in a complex signaling network, and many researchers worry that cancer cells will quickly shift their dependence to a new protein.

And while everyone in the field is excited to see the long-sought clinical milestone, veteran KRAS researchers also are cautious to set expectations for the impact of the drugs and the road ahead.

“This is the start of a journey,” says Frank McCormick, a professor at the University of California, San Francisco, Helen Diller Family Comprehensive Cancer Center and a central figure in the RAS field, which inclues KRAS and its family members HRAS and NRAS. “People may be disappointed initially by the relapse rates, but combination therapy, better dosing, and an understanding of these drugs will push out their benefits significantly.” Getting to that point could take years, McCormick cautions, but “over time, RAS inhibitors will definitely have a place—and will provide major clinical benefit.”

And even if this first wave of drugs targeting KRAS G12C is effective, that still leaves a large swath of cancers driven by other KRAS mutations—around 85% of them, points out Darryl McConnell, research head for Boehringer Ingelheim’s site in Vienna, who is leading an effort to drug KRAS. “I’ve got this pie chart in my head starting with G12D and V and C, but also G13D, G12R/A, and Q61H—we need to drug all of those,” McConnell says. “There’s still a massive amount of work before we are finished with KRAS.”

Credit: Katherine Cohen/Boston Children's Hospital

Timothy Yu, Mila Makovec, and her mother, Julia Vitarello, at Boston Children’s Hospital

Two years ago, Timothy Yu was not a seasoned drug developer. The Boston Children’s Hospital geneticist planned on devoting his career to diagnosing genetic diseases, not treating them.

That all changed with a young girl named Mila Makovec.

Yu’s team pinpointed the root cause of Mila’s neurological condition: a genetic mutation that no one had ever seen before. Yu then realized that an antisense oligonucleotide—a short stretch of synthetic DNA—could potentially cover up the offending mutation and treat her disease.

What followed was an unprecedented sprint to design, test, and manufacture a custom drug that was ultimately approved for injection into Mila—all in the span of 10 months. It’s what the field is calling an N-of-1 therapy, and it shattered records for speed and personalization in drug development.

“Tim Yu showed that someone who has never done an antisense oligonucleotide experiment in their life can rapidly develop a customized drug and treat a patient,” says Arthur Krieg, chief scientific officer of the oligonucleotide company Checkmate Pharmaceuticals. “That’s the huge breakthrough that the field has been waiting for. It’s like the first time someone ran a 4-minute mile,” he adds.

And this breakthrough won’t be the last.

Indeed, Yu is spearheading several more N-of-1 oligonucleotide treatments for rare neurological diseases, including two that may be ready for injection this year or early next. “We’ve been contacted by hundreds of families,” Yu says. He can’t help them all, but his work is invigorating several others to pursue N-of-1 therapies.

For instance, this summer, Columbia University physician Neil Shneider began giving an N-of-1 oligonucleotide to Jaci Hermstad, a woman with a rare form of amyotrophic lateral sclerosis. Although it wasn’t designed just for her—it could help other people with the same mutation—the US Food and Drug Administration approved its use just for Hermstad.

Rohan Seth, an engineer in San Francisco, founded the Lydian Accelerator to develop a custom oligonucleotide for his daughter, Lydia, who was born with a rare genetic form of epilepsy. Seth hopes to create a blueprint based on his experience that other families trying to make N-of-1 therapies can use.

And Rich Horgan, a recent graduate of Harvard Business School, founded a nonprofit called Cure Rare Disease to develop customized CRISPR gene-editing therapies for his younger brother, Terry, and other people with rare forms of muscular dystrophy that oligonucleotides can’t help.

“That so many parents and families are deciding to go it alone shows that the system isn’t working for everybody,” Yu says. “It clearly speaks to this huge, pent-up, unmet need.”

At its annual meeting in October, the Oligonucleotide Therapeutics Society formed a task force to create guidelines for N-of-1 projects. A small number of universities and medical centers are laying foundations for their own N-of-1 oligonucleotide programs. And at least one biotech start-up is quietly working on a business plan to provide the full gamut of services, from genetic diagnosis to N-of-1 drug design.

Many economic, ethical, and regulatory hurdles remain, but Yu showed that N-of-1 drug development is possible. “It is the start of something that will become much more common,” Krieg says. “And the question in my mind isn’t whether there will be tens or hundreds more, but how long will it take us until there are thousands and tens of thousands of Milas?”

CRISPR gene editing has barely graced the clinic. This year, it was tested in a handful of humans in clinical trials for cancer, HIV, and sickle cell disease. And although drug companies have only just begun to offer the first evidence that CRISPR works in humans, scientists are still racing to discover and design new versions.

The classic form of the gene editor that debuted in 2012, CRISPR-Cas9, is like a pair of molecular scissors. It cuts DNA to turn genes on or off. Tools called base editors, which were published in 2016 and 2017 and are often compared to pencils, can change a single nucleotide—or letter—of DNA into another. Although base editors have been hailed as CRISPR 2.0, they can make only 4 of the 12 possible letter-to-letter changes.

Now we have a new system known as prime editing. It can make all 12 letter-to-letter changes, as well as add or remove short stretches of DNA. Call it CRISPR 3.0.

Credit: C&EN

New CRISPR prime editors can add or remove short stretches of DNA or swap any letter for another. In these examples, prime editing adds an ATT sequence to DNA, removes a CTA sequence, and swaps one T:A base pair for a G:C pair and another T:A base pair for a C:G pair.

Prime editing was just the latest in a series of gene-editing innovations published this year. In February, Jennifer Doudna’s lab at the University of California, Berkeley, published a detailed study on a genetically and structurally distinct CRISPR enzyme called CasX, which makes staggered breaks during its cutting of the DNA double helix. That could provide an extra foothold for inserting new DNA sequences into the genome. A start-up, Scribe Therapeutics, is engineering improved versions of CasX for gene editing.

This summer, two labs—one led by Samuel Sternberg at Columbia University and another led by Feng Zhang at Broad Institute of MIT and Harvard—separately described how bacteria use Cas proteins to precisely insert a mobile segment of a DNA called a transposon. They also showed that they could hijack the system to insert a new strand of DNA into a programmable location. Both groups have filed patents on their systems, although they’ve been tested only in bacteria so far.

Then, in October, Broad Institute scientist David Liu, whose lab also invented base editors, debuted his prime editors. The tools are a molecular mash-up of a modified Cas9 protein and a reverse transcriptase enzyme. Cas9 looks to one arm of a guide-RNA molecule to determine where to go in the genome, and the reverse transcriptase reads another arm of the guide RNA to determine what change to make—whether that is swapping one letter for another or adding or removing DNA.

Liu compares prime editors to a search-and-replace function on a computer. And before he even published the paper, he formed a new company, Prime Medicine, to commercialize the technology. Prime has already licensed prime editors to Beam Therapeutics, a company that Liu cofounded to develop medicines with base editors.

And beyond gene editing, several start-ups—Korro Bio, Navega Therapeutics, and Shape Therapeutics—launched this year to develop therapies that edit RNA, the short-lived cousin to DNA. The approach provides a new strategy for treating conditions like chronic pain and inflammation, and it alleviates some concerns about the long-term safety of editing DNA.

New tools yield new intellectual property, which can form the basis of new gene-editing start-ups. But more broadly, tools such as prime editing give biologists yet another way to edit genomes and bring them one step closer to changing DNA in any way they please.

After a bumper crop of 59 new drug approvals in 2018, 2019 looked to be a reset year for the US Food and Drug Administration. As of Nov. 25, 41 new therapies had found their way to the market, some offering first-in-class treatment and others launching into an already-competitive landscape.

While 2019 is shaping up to be a more average year for the number of approvals, the small-molecule drugs and biologics that have been approved so far tackle several underserved diseases and add to patient options for others, says Bernard Munos, senior fellow at the think tank FasterCures. About 25% of the drugs are intended to treat neurological disorders, including Parkinson’s disease and spinal muscular atrophy. And another 25%, including an antibody-drug conjugate, treat various cancers.

The breadth of diseases treated and the types of drugs approved signify the FDA’s commitment to innovation, Munos says. “FDA has worked very hard to become a partner in innovation, to really push companies, to encourage companies to be bolder,” he says.

Here are some of the standout approvals of 2019:

Pretomanid

TB Alliance

Pretomanid

Pretomanid was approved in August as part of a combination treatment to combat tuberculosis, which according to the nonprofit TB Alliance is now the deadliest infectious disease in the world. The nitroimidazole antibacterial, developed by the nonprofit, is used with two existing TB treatments. It’s the second drug approved as part of a special program within the 21st Century Cures Act devoted to new treatments for infectious diseases.

Trikafta

Vertex Pharmaceuticals

Ivacaftor

Tezacaftor

 

Elexacaftor

 

Trikafta

 

Trikafta is a triple therapy for cystic fibrosis (CF), a rare genetic disease in which a misfolded ion transporter can’t get to the surface of lung cells, leading to mucus buildup in the lungs. Two of the three drugs in Trikafta, elexacaftor and tezacaftor, help chaperone the ion transporter to the cell surface, and the remaining drug, ivacaftor, helps keep the ion channel open longer than normal. The therapy, approved in October, is expected to help up to 90% of people with CF.

 

Zulresso

Sage Therapeutics

Brexanolone

Zulresso, approved in March, is the first drug to be cleared by the FDA specifically to treat postpartum depression, which affects one in nine women who give birth in the US. The active ingredient, brexanolone, works on GABA receptors in the brain. Zulresso is administered intravenously, and because it comes with a risk of losing consciousness, a new mom has to be monitored during the infusion, which takes about 2.5 days.

CORRECTION
The story on new drug approvals was updated on Dec. 4, 2019, to clarify that Zulresso is administered in a hospital because it can cause loss of consciousness, not because it is administered intravenously.

Credit: Kymera Therapeutics

The first protein degraders (pink) are bifunctional small molecules that bind to a disease-causing protein (green) at one end and an E3 ligase (blue) at the other. The E3 ligase recruits an enzyme that tags the protein for ubiquitination and subsequent breakdown into amino acids in the proteasome.

Small-molecule drugs took new shape in 2019 as a class of compounds called protein degraders moved from the lab to the clinic. The field of protein degradation, in which a drug forces a bad-behaving protein to be sent to the cellular trash compactor, blossomed with three clinical candidates—two from Arvinas and one from Novartis—a cluster of new players, and several major deals.

Most small-molecule drugs work by inhibiting or modulating the activity of a protein. Degraders upend that paradigm by causing a protein to break down altogether. They are typically heterobifunctional small molecules with an end that tethers to a protein of interest and another that binds to an E3 ligase, which in turn recruits an enzyme that tags the protein for disposal by the proteasome.

For researchers who have spent the past several years trying to design these complex compounds, 2019 was the year in which goalposts shifted beyond scientific success and into the commercial realm.

“Success is no longer showing you can degrade ABC protein in XYZ tissues and establish a certain therapeutic index,” says Anil Vasudevan, senior director of discovery platform technologies at AbbVie. Vasudevan has for the past 4 years led protein degradation efforts at AbbVie, where the team has grown in that time from 2 people to north of 50. “In order for this modality to prove its worth, we’re defining success by clinical proof of concept,” he says.

Nello Mainolfi, CEO of the protein degrader–focused biotech Kymera Therapeutics, agrees. As recently as 2 years ago, discussions of protein degradation at conferences centered on the basics of designing the complex molecules. At the time, many researchers had “a very early, kind of naive understanding of the space,” Mainolfi says.

But this year the field turned a corner. The most meaningful advance came at a recent cancer conference, where Arvinas offered very early data related to the safety and pharmacokinetics of the first two protein degraders to be tested in humans: ARV-110, a degrader of the androgen receptor, and ARV-471, which breaks down the estrogen receptor.

Yale University chemical biologist Craig Crews, who founded Arvinas, points to several takeaways from the data on the two compounds. Because protein degraders are—compared with most small-molecule drugs—quite large, “there have been questions over the years regarding their drug-like properties,” Crews says.

Arvinas’s data show that, in fact, ARV-110 and ARV-471 behave a lot like typical small molecules, he notes. The drug candidates offer “dose proportionality,” meaning the more you give a patient, the more compound can be detected floating around the body. Both compounds appear safe and accumulate in concentrations above those that have proved effective at killing tumors in animal models.

As clinical trials push forward, the field is becoming crowded. Several new firms, including Cullgen, Frontier Medicines, and Plexium, launched in 2019 to pursue small molecules that can break down proteins. And big pharma entered the fray through partnerships. For example, Vertex Pharmaceuticals teamed up with Kymera, Bayer with Arvinas, and Gilead Sciences with Nurix Therapeutics.

“We have enough examples out there that we can make these molecules look like drugs or act like drugs. People feel now is really the time to double down,” Mainolfi says.

Engineered cell therapies that seek and destroy cancer are no longer science fiction, and investors can’t seem to get enough of them. By mid-November this year, cell therapy start-ups had announced raising nearly $2 billion from venture capital firms.

The windfall is easily traced back to the approval of cell therapies from Novartis and Gilead Sciences 2 years ago. “Until 2017, there were a lot of skeptics,” says Usman “Oz” Azam, the former head of cell and gene therapy at Novartis. “What changed everything was the fact that a product got approved.”

The two products are CAR T-cell therapies, in which a person’s own T cells are isolated, genetically engineered, and reinfused to kill cancer. With two success stories in hand, biotech start-ups are eager to improve on and expand what CAR T-cell (CAR-T) therapy, and other kinds of cell therapies, can do.

Azam is leading one such effort as CEO of Tmunity Therapeutics, a start-up that raised $75 million this year. The two approved CAR-T therapies can treat only certain kinds of blood cancer. Firms such as Tmunity are using tools like CRISPR gene editing to impart multiple changes to cell therapies, making them better at spotting and killing solid tumors such as glioblastoma, melanoma, and pancreatic cancer.

The commercial CAR-T therapies must be made from a patient’s own cells. Some start-ups are hoping to fix that limitation by designing allogeneic—also called off-the-shelf—cell therapies made from a stem cell line. That will require additional genetic engineering to ensure that the cells don’t cause an immune reaction in the recipient. This summer, Century Therapeutics launched with $250 million dedicated to the task.

In the fight against cancer, a number of firms are recruiting additional immune cells, such as natural killer cells, regulatory T cells, and γδ T cells. Cell therapy excitement is extending beyond cancer too. For instance, Cabaletta Bio, which raised $50 million this year, is developing T-cell therapies that hunt down and kill the rogue B cells that cause autoimmune diseases.

Investors are also renewing their interest in nonengineered cell therapies for narrow applications. For example, AlloVir is using T cells to treat viral infections in people who recently received bone-marrow transplants. Talaris Therapeutics is using a stem- and immune-cell therapy to prevent immune rejection in people receiving organ transplants. AlloVir raised $120 million and Talaris raised $100 million in 2019.

“The trend has been to larger and larger raises,” says Janet Lambert, CEO of the Alliance for Regenerative Medicine, an industry group for cell and gene therapy. That can be chalked up partially to investor enthusiasm, she says, but meeting the challenge of manufacturing cell therapies also demands robust funding.

“If you can’t make the product up front, then you shouldn’t be in the business,” Azam says. That’s forcing young companies to build out their manufacturing capacity in parallel with preclinical and clinical experiments.

Lambert expects investment to continue apace in 2020. “On the start-up side, I think it is going to be another bang-up year,” she says. “And it’s not just a CAR-T story anymore.”

Credit: Olivia Acland/Reuters/Newscom

Hundreds of thousands of people in the Democratic Republic of Congo have been vaccinated against Ebola since an outbreak was declared in August 2018.

An Ebola outbreak in the Democratic Republic of the Congo (DRC) could have been much more deadly, public health experts say, if not for an experimental vaccine used to treat hundreds of thousands of people. The efficacy data from the vaccine, called Ervebo, helped European regulators decide in November to make it the first approved Ebola vaccine on the market.

Health-care workers began administering Ervebo in August 2018. The single-dose, live vaccine consists of vesicular stomatitis virus engineered to express an Ebola protein. It is more than 90% successful in preventing new infections, according to the World Health Organization. The vaccine was created by Canada’s National Microbiology Laboratory, developed by NewLink Genetics, and licensed by Merck & Co.

More than 223,000 people in the DRC have received the Merck vaccine, despite political instability, medical mistrust, and even violence against health-care workers. About 2,200 people have died of infection.

The success of the vaccination campaign in stemming the spread of Ebola has been remarkable, says Peter Hotez, who runs the National School of Tropical Medicine at Baylor College of Medicine.

“This is an extraordinary story,” Hotez says. “Imagine if they hadn’t been able to do this. If it weren’t for the vaccine, this could potentially have destabilized the whole African continent.”

As the outbreak continues, public health officials want to stop the virus from spreading beyond the most affected regions in the DRC: North Kivu, South Kivu, and Ituri. To that end, Janssen Pharmaceutical, part of Johnson & Johnson, initiated a trial of its own experimental Ebola vaccine in mid-November in Goma, a city in North Kivu on the border with Rwanda, and donated 500,000 doses to health-care facilities.

Daniel Bausch, a tropical-medicine specialist in Goma who is helping lead the Janssen trial, explains that the vaccine features two components delivered in two doses: the first injection contains an adenovirus that expresses an Ebola protein; the second has a modified vaccinia virus engineered to express an Ebola protein and a protein from the similar Marburg virus.

Because the two doses are administered 56 days apart, clinics will be situated in neighborhoods in Goma that health-care workers know are frequented by people from the affected regions. In contrast to the ring vaccination protocol used with the Merck product—after infection is confirmed, that person’s contacts, and contacts’ contacts, are vaccinated—the Janssen trial will vaccinate nearly anyone who comes into the clinics, Bausch says.

With evidence suggesting that the outbreak is slowing, the Janssen trial could fall short of the numbers needed to prove the vaccine’s efficacy. But Bausch still sees plenty to learn from the experience, such as how best to refrigerate and disseminate the vaccine, how to talk and teach about vaccination, and how best to set up the systems to administer it—information that will be useful if Ebola resurges.

“Everyone is cautiously optimistic,” Bausch says about the slowing of the outbreak. “We don’t want to have an outbreak that is ongoing just so we can collect scientific data.”

Credit: Insilico Medicine

Artificial intelligence identified this potential kinase inhibitor drug in 46 days.

Artificial intelligence algorithms can make accurate predictions about unfamiliar systems using data they’ve been exposed to. In the vast and complex space of molecules, reactions, and biological interactions, that makes AI an attractive alternative to human drug hunters. But if AI will lead a revolution in drug discovery, it didn’t happen in 2019.

“We have bits and pieces where people have made really good advances,” says Mike Tarselli, scientific director of the Society for Laboratory Automation and Screening, but he saw no transformative breakthrough.

Among the advances: Researchers at the Massachusetts Institute of Technology showed off a system that uses AI and robotics to plan and execute syntheses of drug molecules, which could help chemists search for and evaluate potential therapies. The researchers behind Synthia, an AI-based retrosynthesis planner, introduced an algorithm that could help drug chemists avoid synthetic routes that are already patent protected. And the start-up Insilico Medicine demonstrated an algorithm that found a potential kinase inhibitor drug in just 46 days, a feat that experts in AI and medicinal chemistry praised with the caveat that the molecule the algorithm found seemed like low-hanging fruit.

“We still don’t have an important therapeutic molecule that was designed by AI and went into clinical trials,” Regina Barzilay, an AI expert at MIT, says. But the field is still young, she adds, pointing to improvements in AI algorithms’ accuracy and the adoption of more standardized and rigorous benchmarks as evidence that the discipline is maturing and improving.

Even without a revolution, AI-powered drug discovery came into its own in 2019, says Timothy Cernak, a medicinal chemist at the University of Michigan who previously worked for Merck & Co. “This wasn’t even a field last year,” Cernak says. Now, he is noticing a deepening of the relationship between AI researchers and organic chemists.

Some companies have decided that AI still isn’t ready for prime time. IBM reportedlystopped marketing its Watson system for use in drug discovery, shifting resources to clinical development. “That should say just how challenging the endeavor of drug design is,” Tarselli says. But others continue to invest in the area, with companies like Insilico and Exscientia announcing large investments.

Elsewhere, groups of pharmaceutical companies and others are pooling resources and knowledge to try to accelerate progress. The Machine Learning Ledger Orchestration for Drug Discovery project wants to help competitors share drug library data so AI algorithms have more information to learn from. And members of MIT’s Machine Learning for Pharmaceutical Discovery and Synthesis Consortium, which Barzilay is part of, are working to develop new AI algorithms and tools.