Posted on by Lawrence Tabak, D.D.S., Ph.D.
Each year, more than 2.8 million people in the United States develop bacterial infections that don’t respond to treatment and sometimes turn life-threatening . Their infections are antibiotic-resistant, meaning the bacteria have changed in ways that allow them to withstand our current widely used arsenal of antibiotics. It’s a serious and growing health-care problem here and around the world. To fight back, doctors desperately need new antibiotics, including novel classes of drugs that bacteria haven’t seen and developed ways to resist.
Developing new antibiotics, however, involves much time, research, and expense. It’s also fraught with false leads. That’s why some researchers have turned to harnessing the predictive power of artificial intelligence (AI) in hopes of selecting the most promising leads faster and with greater precision.
It’s a potentially paradigm-shifting development in drug discovery, and a recent NIH-funded study, published in the journal Molecular Systems Biology, demonstrates AI’s potential to streamline the process of selecting future antibiotics . The results are also a bit sobering. They highlight the current limitations of one promising AI approach, showing that further refinement will still be needed to maximize its predictive capabilities.
These findings come from the lab of James Collins, Massachusetts Institute of Technology (MIT), Cambridge, and his recently launched Antibiotics-AI Project. His audacious goal is to develop seven new classes of antibiotics to treat seven of the world’s deadliest bacterial pathogens in just seven years. What makes this project so bold is that only two new classes of antibiotics have reached the market in the last 50 years!
In the latest study, Collins and his team looked to an AI program called AlphaFold2 . The name might ring a bell. AlphaFold’s AI-powered ability to predict protein structures was a finalist in Science Magazine’s 2020 Breakthrough of the Year. In fact, AlphaFold has been used already to predict the structures of more than 200 million proteins, or almost every known protein on the planet .
AlphaFold employs a deep learning approach that can predict most protein structures from their amino acid sequences about as well as more costly and time-consuming protein-mapping techniques.
In the deep learning models used to predict protein structure, computers are “trained” on existing data. As computers “learn” to understand complex relationships within the training material, they develop a model that can then be applied for making predictions of 3D protein structures from linear amino acid sequences without relying on new experiments in the lab.
Collins and his team hoped to combine AlphaFold with computer simulations commonly used in drug discovery as a way to predict interactions between essential bacterial proteins and antibacterial compounds. If it worked, researchers could then conduct virtual rapid screens of millions of new synthetic drug compounds targeting key bacterial proteins that existing antibiotics don’t. It would also enable the rapid development of antibiotics that work in novel ways, exactly what doctors need to treat antibiotic-resistant infections.
To test the strategy, Collins and his team focused first on the predicted structures of 296 essential proteins from the Escherichia coli bacterium as well as 218 antibacterial compounds. Their computer simulations then predicted how strongly any two molecules (essential protein and antibacterial) would bind together based on their shapes and physical properties.
It turned out that screening many antibacterial compounds against many potential targets in E. coli led to inaccurate predictions. For example, when comparing their computational predictions with actual interactions for 12 essential proteins measured in the lab, they found that their simulated model had about a 50:50 chance of being right. In other words, it couldn’t identify true interactions between drugs and proteins any better than random guessing.
They suspect one reason for their model’s poor performance is that the protein structures used to train the computer are fixed, not flexible and shifting physical configurations as happens in real life. To improve their success rate, they ran their predictions through additional machine-learning models that had been trained on data to help them “learn” how proteins and other molecules reconfigure themselves and interact. While this souped-up model got somewhat better results, the researchers report that they still aren’t good enough to identify promising new drugs and their protein targets.
What now? In future studies, the Collins lab will continue to incorporate and train the computers on even more biochemical and biophysical data to help with the predictive process. That’s why this study should be interpreted as an interim progress report on an area of science that will only get better with time.
But it’s also a sobering reminder that the quest to find new classes of antibiotics won’t be easy—even when aided by powerful AI approaches. We certainly aren’t there yet, but I’m confident that we will get there to give doctors new therapeutic weapons and turn back the rise in antibiotic-resistant infections.
 2019 Antibiotic resistance threats report. Centers for Disease Control and Prevention.
 Benchmarking AlphaFold-enabled molecular docking predictions for antibiotic discovery. Wong F, Krishnan A, Zheng EJ, Stark H, Manson AL, Earl AM, Jaakkola T, Collins JJ. Molecular Systems Biology. 2022 Sept 6. 18: e11081.
 Highly accurate protein structure prediction with AlphaFold. Jumper J, Evans R, Pritzel A, Kavukcuoglu K, Kohli P, Hassabis D., et al. Nature. 2021 Aug;596(7873):583-589.
 ‘The entire protein universe’: AI predicts shape of nearly every known protein. Callaway E. Nature. 2022 Aug;608(7921):15-16.
Antimicrobial (Drug) Resistance (National Institute of Allergy and Infectious Diseases/NIH)
Collins Lab (Massachusetts Institute of Technology, Cambridge)
AlphaFold (Deep Mind, London, United Kingdom)
NIH Support: National Institute of Allergy and Infectious Diseases; National Institute of General Medical Sciences
Posted on by Dr. Francis Collins
Researchers continue to make progress with cancer immunotherapy, a type of treatment that harnesses the body’s own immune cells to attack cancer. But Kole Roybal wants to help move the field further ahead by engineering patients’ immune cells to detect an even broader range of cancers and then launch customized attacks against them.
With an eye toward developing the next generation of cell-based immunotherapies, this synthetic biologist at University of California, San Francisco, has already innovatively hacked into how certain cells communicate with each other. Now, he and his research team are using a 2018 NIH Director’s New Innovator Award to build upon that progress.
Roybal’s initial inspiration is CAR-T therapy, one of the most advanced immunotherapies to date. In CAR-T therapy, some of a cancer patient’s key immune cells, called T cells, are removed and engineered in a way that they begin to produce new surface proteins called chimeric antigen receptors (CARs). Those receptors allow the cells to recognize and attack cancer cells more effectively. After expanding the number of these engineered T cells in the lab, doctors infuse them back into patients to enhance their immune systems’s ability to seek-and-destroy their cancer.
As helpful as this approach has been for some people with leukemia, lymphoma, and certain other cancers, it has its limitations. For one, CAR-T therapy relies solely on a T cell’s natural activation program, which can be toxic to patients if the immune cells damage healthy tissues. In other patients, the response simply isn’t strong enough to eradicate a cancer.
Roybal realized that redirecting T cells to attack a broader range of cancers would take more than simply engineering the receptors to bind to cancer cells. It also would require sculpting novel immune cell responses once those receptors were triggered.
Roybal found a solution in a new class of lab-made receptors known as Synthetic Notch, or SynNotch, that he and his colleagues have been developing over the last several years [1, 2]. Notch protein receptors play an essential role in developmental pathways and cell-to-cell communication across a wide range of animal species. What Roybal and his colleagues found especially intriguing is the protein receptors’ mode of action is remarkably direct.
When a protein binds the Notch receptor, a portion of the receptor breaks off and heads for the cell nucleus, where it acts as a switch to turn on other genes. They realized that engineering a cancer patient’s immune cells with synthetic SynNotch receptors could offer extraordinary flexibility in customized sensing and response behaviors. What’s more, the receptors could be tailored to respond to a number of user-specified cues outside of a cell.
In his NIH-supported work, Roybal will devise various versions of SynNotch-engineered cells targeting solid tumors that have proven difficult to treat with current cell therapies. He reports that they are currently developing the tools to engineer cells to sense a broad spectrum of cancers, including melanoma, glioblastoma, and pancreatic cancer.
They’re also engineering cells equipped to respond to a tumor by producing a range of immune factors, including antibodies known to unleash the immune system against cancer. He says he’ll also work on adding engineered SynNotch molecules to other immune cell types, not just T cells.
Given the versatility of the approach, Roybal doesn’t plan to stop there. He’s also interested in regenerative medicine and in engineering therapeutic cells to treat autoimmune conditions. I’m looking forward to see just how far these and other next-gen cell therapies will take us.
 Engineering Customized Cell Sensing and Response Behaviors Using Synthetic Notch Receptors. Morsut L, Roybal KT, Xiong X, Gordley RM, Coyle SM, Thomson M, Lim WA. Cell. 2016 Feb 11;164(4):780-91.
 Engineering T Cells with Customized Therapeutic Response Programs Using Synthetic Notch Receptors. Roybal KT, Williams JZ, Morsut L, Rupp LJ, Kolinko I, Choe JH, Walker WJ, McNally KA, Lim WA. Cell. 2016 Oct 6;167(2):419-432.e16.
Car-T Cells: Engineering Patients’ Immune Cells to Treat Cancers (National Cancer Institute/NIH)
Synthetic Biology for Technology Development (National Institute of Biomedical Imaging and Bioengineering/NIH)
Roybal Lab (University of California, San Francisco)
Roybal Project Information (NIH RePORTER)
NIH Support: Common Fund; National Cancer Institute
Posted on by Dr. Francis Collins
For Salmonella and many other disease-causing bacteria that find their way into our bodies, infection begins with a poke. That’s because these bad bugs are equipped with a needle-like protein filament that punctures the outer membrane of human cells and then, like a syringe, injects dozens of toxic proteins that help them replicate.
Cammie Lesser at Massachusetts General Hospital and Harvard Medical School, Cambridge, and her colleagues are now on a mission to bioengineer strains of bacteria that don’t cause disease to make these same syringes, called type III secretion systems. The goal is to use such “good” bacteria to deliver therapeutic molecules, rather than toxins, to human cells. Their first target is the gastrointestinal tract, where they hope to knock out hard-to-beat bacterial infections or to relieve the chronic inflammation that comes with inflammatory bowel disease (IBD).
Posted on by Dr. Francis Collins
The term “freeze-dried” may bring to mind those handy MREs (Meals Ready to Eat) consumed by legions of soldiers, astronauts, and outdoor adventurers. But if one young innovator has his way, a test that features freeze-dried biosensors may soon be a key ally in our nation’s ongoing campaign against the very serious threat of antibiotic-resistant bacterial infections.
Each year, antibiotic-resistant infections account for more than 23,000 deaths in the United States. To help tackle this challenge, Ahmad (Mo) Khalil, a researcher at Boston University, recently received an NIH Director’s New Innovator Award to develop a system that can more quickly determine whether a patient’s bacterial infection will respond best to antibiotic X or antibiotic Y—or, if the infection is actually viral rather than bacterial, no antibiotics are needed at all.