The team - co-led by the University of Michigan, the Centre for Cooperative Research in Biomaterials-CIC biomaGUNE, the University of Vigo, and the University of Antwerp - has received €9.3m from the European Research Council to conduct the research.
They aim to design a machine-learning model that can predict which nanoparticle structures will bind to specific biomolecules, such as proteins on the cell walls of pathogens.
Selective binding to various proteins and lipids is useful in detecting and treating diseases; nanoparticles designed to twist like their protein targets have already been used to detect molecular signs of Parkinson's disease and lung cancer in laboratory studies.
The chirality of the nanoparticles determines how strongly they bind with certain proteins and impacts whether they absorb or twist specific colours and orientations of light. Researchers can detect the change to confirm the presence of the target protein. Light can also be used to selectively heat the protein, destroying harmful cells attached to the nanoparticles.
Designing chiral nanoparticles that can bind to specific biological targets remains a challenge.
"We have to understand when the nanoparticles form strong complexes with specific proteins like a lock and key, but predicting those complexes is hard," said Nick Kotov, the Irving Langmuir Distinguished University Professor of Chemical Sciences and Engineering at the University of Michigan and a co-principal investigator on the project.
"Many biomolecules in the body could interact with simple, spherical nanoparticles, but these interactions are nonspecific. We need to make nanoparticles with more complex shapes across different scales that recognise and bind to biological targets with high specificity."
MORE FROM MEDICAL & HEALTHCARE
According to the University of Michigan, it can take months or years of experiments to test the interactions in every relevant environment, so the researchers hope that their AI tool will speed up nanoparticle engineering by reducing trial and error.
"We hope to offer the scientific community a tool to produce nanoparticles on demand and obtain artificially manufactured materials with properties never seen before," said Luis Liz-Marzán, a professor at CIC biomaGUNE and the University of Vigo and a co-principal investigator.
Beyond medicine, combinations of nanoparticles and proteins could be used to build devices that can guide and change light signals, enabling faster data transfer between nearby users within sight.
To build the AI-powered nanoparticle generator, the researchers will first make a training dataset using AlphaFold, an AI tool that predicts protein shapes. With the initial data, the researcher's model will predict protein shapes that fit together and how nanoparticles could be designed with similar shapes.
The researchers will verify model predictions by creating those matching nanoparticles and testing how the nanoparticles interact with their target proteins in lab experiments. Differences between the experimental results and the model's predictions will then be used to refine the model.
The team will also need a training dataset of 3D images of their nanoparticles, which are conventionally made with transmission electron tomography. It can take a day to collect all the images required for a 3D model of a single nanoparticle and, because of variations among the nanoparticles in any single batch, hundreds of 3D models need to be made.
Sara Bals, a professor at the University of Antwerp's EMAT electron microscopy centre and a co-principal investigator, said nano researchers have so far overlooked a particular phenomenon, namely that when the microscope's electron beam passes through an object, electrons in the object's surface escape, leaving behind holes. Other electrons will move to fill those holes, generating an electrical current shaped by the object's topography. That current can then be translated into a 3D image.
"These interactions have always been there, but nobody has thought to use them for creating 3D images of nanoparticles," Bals said in a statement. "With this technique, we can investigate hundreds of nanoparticles in a single image."
Heat network to set bar for decarbonising heat
The report is somewhat outdated and seems to give emphasis to hydrogen and carbon capture. Thermal energy sources do not include those for electricity...