AI discovery of a new way to enter a cell could boost molecular design

Serendipity.

Happy accidents have long helped scientists discover new materials. And it’s serendipitous insights from physical modelling and AI-boosted computer simulation that have led us to an unexpected finding that could create more opportunities for molecular design.

All forms of life rely on cellular membrane as a biological defense mechanism against bacterial and viral pathogens, and membrane disruption typically happens through circular pores. Having studied how molecules disrupt or permeate a cell’s membrane, our team of researchers from IBM, Oxford, Cambridge and the National Physical Laboratory have discovered a fundamentally new, extremely sensitive, mechanism of biological membrane disruption.

In our paper, “Switching Cytolytic Nanopores into Antimicrobial Fractal Ruptures by a Single Side Chain Mutation,”¹ published in ACS Nano and selected for the cover (pictured), we describe how AI-designed antimicrobial peptides interact with computational models of a cellular membrane. We discovered that a carefully chosen but minimal chemical modification (a single mutation in the amino acid sequence) alters the nature of the interaction of the peptide with the membrane—both penetration depth and orientation. Amino acids are the body’s basic building blocks that form chains—peptides and proteins.

In lab-based experiments using cutting-edge imaging at the nanoscale, we found that this mutation has unexpectedly large effects. It switches the molecular mode of action from conventional circular pore formation to a previously unknown fractal rupture mode that still shows strong antimicrobial potency, or effectiveness. While researching properties of turbulence over telephone lines at the IBM Thomas J. Watson Research Center in Yorktown Heights, NY, Benoit Mandelbrot discovered the principles that would later form the new field of fractal geometry. This discovery made it possible, for the first time in the history of mankind, to describe nature with math. Read more about fractals and the Mandelbrot set.Fractals are self-similar patterns frequently occurring in nature—for example, in snowflakes and Romanesco broccoli—but until now haven’t been observed in cellular membranes, simulated or real.

Our results show that minimal chemical changes can have very large functional consequences. Understanding how to control these processes can help us tune material properties from the molecular scale upwards.

Disrupting the membrane

Biological membranes are highly complex systems with many chemical components and processes occurring at various lengths and timescales. Researchers have long studied molecular mechanisms of different membrane disruption phenomena in peptides, proteins, antibiotics and nanoparticles, but so far haven’t found a general or unifying mechanism.

Intriguingly, we weren’t looking for such a mechanism at all.

Our assumption was that the successful candidates would behave as usual, disrupting the membrane by forming a circular pore on its surface.

Having designed a series of very small peptides, we ran the initial simulations. We found that the molecules seemed to function as potent antimicrobials, able to enter our simulated cell membranes. They assembled in the membrane to disrupt the its structure at the molecular level. But how?

The answer was in the amino acids themselves

We found that when we swapped one aminoacid (alanine) for another one (lysine), the mutation led to the membrane’s rupture—and a never-before seen stunning fractal pattern.

This work builds on IBM Research’s ongoing efforts to accelerate the pace at which new drugs and therapies can be discovered, tested and brought to market. Recently, IBM researchers also published work in Nature Biomedical Engineering² demonstrating a generative AI system that can help to speed up the design of molecules for novel antibiotics, outperforming other design methods by nearly 10 percent. This model has already created two new, non-toxic antimicrobial peptides (AMP) with strong broad-spectrum potency.

But simulation isn’t life. Once synthetized in the lab, some of these molecules did indeed disrupt a real cell’s membrane and produced a fractal rupture design. Almost immediately, the anomalous fractal rupture pathway was converted into conventional, circular, pore formation.

The chemically cued instability mode we’ve discovered appears distinct from any known manifestations of membrane organization, phase separation, pore formation phenomena or other forms of segregation. While the mechanism has entirely new properties and new behavior, the biological relevance and exploitation opportunities of some of these findings are still being explored. Still, the discovery of a fundamentally new instability mode in a biological membrane has potentially wide implications for drug discovery and drug delivery.