Skip to main content

E. coli

Using AI to Find New Antibiotics Still a Work in Progress

Posted on by

Protein over a computer network

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 [1]. 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 [2]. 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 [3]. 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 [4].

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.

References:

[1] 2019 Antibiotic resistance threats report. Centers for Disease Control and Prevention.

[2] 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.

[3] 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.

[4] ‘The entire protein universe’: AI predicts shape of nearly every known protein. Callaway E. Nature. 2022 Aug;608(7921):15-16.

Links:

Antimicrobial (Drug) Resistance (National Institute of Allergy and Infectious Diseases/NIH)

Collins Lab (Massachusetts Institute of Technology, Cambridge)

The Antibiotics-AI Project, The Audacious Project (TED)

AlphaFold (Deep Mind, London, United Kingdom)

NIH Support: National Institute of Allergy and Infectious Diseases; National Institute of General Medical Sciences


Using R2D2 to Understand RNA Folding

Posted on by

If you love learning more about biology at a fundamental level, I have a great video for you! It simulates the 3D folding of RNA. RNA is a single stranded molecule, but it is still capable of forming internal loops that can be stabilized by base pairing, just like its famously double-stranded parent, DNA. Understanding more about RNA folding may be valuable in many different areas of biomedical research, including developing ways to help people with RNA-related diseases, such as certain cancers and neuromuscular disorders, and designing better mRNA vaccines against infectious disease threats (like COVID-19).

Because RNA folding starts even while an RNA is still being made in the cell, the process has proven hugely challenging to follow closely. An innovative solution, shown in this video, comes from the labs of NIH grantees Julius Lucks, Northwestern University, Evanston, IL, and Alan Chen, State University of New York at Albany. The team, led by graduate student Angela Yu and including several diehard Star Wars fans, realized that to visualize RNA folding they needed a technology platform that, like a Star Wars droid, is able to “see” things that others can’t. So, they created R2D2, which is short for Reconstructing RNA Dynamics from Data.

What’s so groundbreaking about the R2D2 approach, which was published recently in Molecular Cell, is that it combines experimental data on RNA folding at the nucleotide level with predictive algorithms at the atomic level to simulate RNA folding in ultra-slow motion [1]. While other computer simulations have been available for decades, they have lacked much-needed experimental data of this complex folding process to confirm their mathematical modeling.

As a gene is transcribed into RNA one building block, or nucleotide, at a time, the elongating RNA strand folds immediately before the whole molecule is fully assembled. But such folding can create a problem: the new strand can tie itself up into a knot-like structure that’s incompatible with the shape it needs to function in a cell.

To slip this knot, the cell has evolved immediate corrective pathways, or countermoves. In this R2D2 video, you can see one countermove called a toehold-mediated strand displacement. In this example, the maneuver is performed by an ancient molecule called a single recognition particle (SRP) RNA. Though SRP RNAs are found in all forms of life, this one comes from the bacterium Escherichia coli and is made up of 114 nucleotides.

The colors in this video highlight different domains of the RNA molecule, all at different stages in the folding process. Some (orange, turquoise) have already folded properly, while another domain (dark purple) is temporarily knotted. For this knotted domain to slip its knot, about 5 seconds into the video, another newly forming region (fuchsia) wiggles down to gain a “toehold.” About 9 seconds in, the temporarily knotted domain untangles and unwinds, and, finally, at about 23 seconds, the strand starts to get reconfigured into the shape it needs to do its job in the cell.

Why would evolution favor such a seemingly inefficient folding process? Well, it might not be inefficient as it first appears. In fact, as Chen noted, some nanotechnologists previously invented toehold displacement as a design principle for generating synthetic DNA and RNA circuits. Little did they know that nature may have scooped them many millennia ago!

Reference:

[1] Computationally reconstructing cotranscriptional RNA folding from experimental data reveals rearrangement of non-naïve folding intermediates. Yu AM, Gasper PM Cheng L, Chen AA, Lucks JB, et. al. Molecular Cell 8, 1-14. 18 February 2021.

Links:

Ribonucleic Acid (RNA) (National Human Genome Research Institute/NIH)

Chen Lab (State University of New York at Albany)

Lucks Laboratory (Northwestern University, Evanston IL)

NIH Support: National Institute of General Medical Sciences; Common Fund


Some ‘Hospital-Acquired’ Infections Traced to Patient’s Own Microbiome

Posted on by

Bacteria in both blood and gut

Caption: New computational tool determines whether a gut microbe is the source of a hospital-acquired bloodstream infection
Credit: Fiona Tamburini, Stanford University, Palo Alto, CA

While being cared for in the hospital, a disturbingly large number of people develop potentially life-threatening bloodstream infections. It’s been thought that most of the blame lies with microbes lurking on medical equipment, health-care professionals, or other patients and visitors. And certainly that is often true. But now an NIH-funded team has discovered that a significant fraction of these “hospital-acquired” infections may actually stem from a quite different source: the patient’s own body.

In a study of 30 bone-marrow transplant patients suffering from bloodstream infections, researchers used a newly developed computational tool called StrainSifter to match microbial DNA from close to one-third of the infections to bugs already living in the patients’ large intestines [1]. In contrast, the researchers found little DNA evidence to support the notion that such microbes were being passed around among patients.


A Lean, Mean DNA Packaging Machine

Posted on by

Three views of bacteriophage T4

Credit: Victor Padilla-Sanchez, The Catholic University of America, Washington, D.C.

All plants and animals are susceptible to viral infections. But did you know that’s also true for bacteria? They get nailed by viruses called bacteriophages, and there are thousands of them in nature including this one that resembles a lunar lander: bacteriophage T4 (left panel). It’s a popular model organism that researchers have studied for nearly a century, helping them over the years to learn more about biochemistry, genetics, and molecular biology [1].

The bacteriophage T4 infects the bacterium Escherichia coli, which normally inhabits the gastrointestinal tract of humans. T4’s invasion starts by touching down on the bacterial cell wall and injecting viral DNA through its tube-like tail (purple) into the cell. A DNA “packaging machine” (middle and right panels) between the bacteriophage’s “head” and “tail” (green, yellow, blue spikes) keeps the double-stranded DNA (middle panel, red) at the ready. All the vivid colors you see in the images help to distinguish between the various proteins or protein subunits that make up the intricate structure of the bacteriophage and its DNA packaging machine.


Adding Letters to the DNA Alphabet

Posted on by

semi-synthetic bacterium

Credit: William B. Kiosses

The recipes for life, going back billions of years to the earliest single-celled organisms, are encoded in a DNA alphabet of just four letters. But is four as high as the DNA code can go? Or, as researchers have long wondered, is it chemically and biologically possible to expand the DNA code by a couple of letters?

A team of NIH-funded researchers is now answering these provocative questions. The researchers recently engineered a semi-synthetic bacterium containing DNA with six letters, including two extra nucleotides [1, 2]. Now, in a report published in Nature, they’ve taken the next critical step [3]. They show that bacteria, like those in the photo, are not only capable of reliably passing on to the next generation a DNA code of six letters, they can use that expanded genetic information to produce novel proteins unlike any found in nature.


Next Page