Rapid classification of β-lactamase activity through AI-based structure prediction and electric field analysis

Bacteria have been fighting back against our medicines for decades, evolving tiny molecular machines that can dismantle antibiotics before they ever reach their target. Among the most dangerous of these defenders are enzymes called beta-lactamases, which act like molecular scissors, snipping the chemical bonds of a major class of drugs known as beta-lactams. When these drugs fail, common infections can become deadly, and the few remaining "last-resort" antibiotics, such as carbapenems, become useless. The challenge for scientists is that bacteria mutate rapidly, creating new versions of these enzymes that can cut through even the strongest drugs. To stay ahead, researchers need a way to predict how well a new bacterial variant will destroy a drug, ideally just by looking at its genetic code, without waiting for slow and expensive laboratory experiments.

A team of researchers has now developed a fast, computer-based method to predict exactly how efficiently these bacterial enzymes can break down carbapenem antibiotics. Instead of waiting to grow bacteria in a lab, they used artificial intelligence to build a 3D model of the enzyme holding onto the drug, creating a snapshot of the moment just before the drug is destroyed. From this snapshot, they calculated the invisible electric forces inside the enzyme's active site. They found that the strength and direction of these electric fields act as a reliable fingerprint for the enzyme's power. If the electric field is strong enough in the right direction, the enzyme will likely destroy the drug; if not, the drug will survive. This approach allows scientists to scan new bacterial sequences and immediately estimate their resistance levels, offering a crucial tool for tracking the spread of superbugs.

The story of antibiotic resistance is a race between human ingenuity and bacterial evolution. Beta-lactam antibiotics, which include penicillin and its modern relatives, work by jamming the machinery bacteria use to build their cell walls. However, many bacteria produce beta-lactamase enzymes that recognize these drugs and cut them open, rendering them harmless. The situation is particularly dire with carbapenems, a powerful class of antibiotics often reserved for the most severe infections. When bacteria evolve enzymes that can also cut carbapenems, treatment options vanish, leading to higher rates of illness and death. Traditionally, figuring out whether a specific bacterial strain can resist these drugs requires growing the bacteria, isolating the enzyme, and testing it in a lab. This process is slow and cannot keep pace with the speed at which new mutations appear.

To solve this, the researchers turned to a different kind of tool: a computational protocol that relies on the laws of physics and the power of artificial intelligence. They started with the genetic sequence of ten different types of beta-lactamase enzymes, some of which are known to destroy carbapenems and others that are not. Using an AI system trained on known protein structures, they predicted what the 3D shape of each enzyme would look like when it was holding onto a carbapenem drug molecule. This created a virtual "acyl-enzyme complex," a snapshot of the enzyme and drug locked together, ready for the next step of the reaction.

The critical moment in this process is called deacylation, where a water molecule attacks the bond holding the drug to the enzyme, breaking it apart. The researchers focused on the very first instant of this attack, a fleeting moment known as the transition state. In a full simulation, capturing this moment requires immense computing power and time. Instead, the team used a clever shortcut. They generated a small set of "approximate transition state" structures by gently nudging the virtual water molecule toward the drug and letting the system settle into a stable position. This required only a tiny fraction of the computing time usually needed, yet it preserved the essential physical details of the reaction.

Once they had these virtual snapshots, the team measured the electric field inside the enzyme's active site. An electric field is an invisible force that can push or pull on charged particles. In this case, the researchers were interested in how the enzyme's electric field interacted with a specific part of the drug molecule, the carbonyl group, which carries a negative charge. They found that the strength of this electric field, specifically how well it stabilized the drug at the moment of attack, was directly linked to how fast the enzyme could destroy the drug.

The results were striking. When the researchers compared their calculated electric field values with real-world data from laboratory experiments, the numbers matched with high precision. Enzymes that are known to be powerful carbapenem destroyers showed a strong, negative electric field that perfectly stabilized the drug for destruction. Enzymes that cannot destroy the drug showed a much weaker field. The correlation was so strong that the electric field value alone could distinguish between enzymes that break down carbapenems and those that do not. This held true even when the researchers used very short simulation times, suggesting the method is both fast and reliable.

One of the most important findings was that the researchers did not need to know the exact 3D structure of the enzyme beforehand. By starting with just the genetic sequence and using AI to predict the shape, they could still get accurate results. This means the method can be applied to brand-new bacterial variants as soon as their genetic code is sequenced, without waiting for a crystal structure to be solved in a lab. The team also discovered that the specific orientation of a small part of the drug molecule, a hydroxyethyl group, influenced the results. By focusing on the electric field generated by the enzyme's core structure and ignoring this flexible part, they improved the accuracy of their predictions even further.

The study also highlighted the role of water molecules inside the enzyme. While the protein structure itself provides a strong electric field, the water molecules surrounding the reaction site contribute significantly to the enzyme's power. The researchers found that including these water molecules in their calculations was essential for matching the experimental data. This suggests that the environment inside the enzyme is just as important as the enzyme's shape itself.

This work does not claim to have solved the problem of antibiotic resistance, but it offers a powerful new lens through which to view it. By showing that a simple calculation of electric forces can predict complex biological activity, the researchers have provided a tool that is both fast and grounded in physical reality. The method successfully separated the dangerous carbapenem-destroying enzymes from the harmless ones, confirming that the ability to break down these drugs is rooted in the specific electrical landscape of the enzyme's active site.

The implications for public health are significant. As new bacterial strains emerge, scientists can now use this protocol to quickly assess their threat level. Instead of waiting weeks for lab tests, they could potentially analyze the genetic code of a new variant and immediately know if it poses a risk to carbapenem treatment. This speed could help doctors choose the right antibiotics sooner and guide the development of new drugs that are designed to resist these electric fields. The study demonstrates that by understanding the invisible forces at work inside these molecular machines, we can better predict how they will behave and, ultimately, how to stop them.