Building Artificial Neurons with Mathematics

Credit: Pixabay/CC0 Public Domain by Ecole Polytechnique Federale de Lausanne

EPFL’s Blue Brain Project has found a way to use only mathematics to automatically draw neurons in 3D, meaning we are getting closer to being able to build digital twins of brains.

Santiago Ramón y Cajal, a Spanish physician from the turn of the 19th century, is considered by most to be the father of modern neuroscience. He stared down a microscope day and night for years, fascinated by chemically stained neurons he found in slices of human brain tissue.

By hand, he painstakingly drew virtually every new type of neuron he came across using nothing more than pen and paper. As the Charles Darwin for the brain, he mapped every detail of the forest of neurons that make up the brain, calling them the “butterflies of the brain.”

Today, 200 years later, Blue Brain has found a way to dispense with the human eye, pen and paper, and use only mathematics to automatically draw neurons in 3D as digital twins. Math can now be used to capture all the “butterflies of the brain,” which allows us to use computers to build any and all the billons of neurons that make up the brain. And that means we are getting closer to being able to build digital twins of brains.

These billions of neurons form trillions of synapses—where neurons communicate with each other. Such complexity needs comprehensive neuron models and accurately reconstructed detailed brain networks in order to replicate the healthy and disease states of the brain. Efforts to build such models and networks have historically been hampered by the lack of experimental data available.

But now, scientists at the EPFL Blue Brain Project using algebraic topology, a field of Math, have created an algorithm that requires only a few examples to generate large numbers of unique cells. Using this algorithm—the Topological Neuronal Synthesis (TNS), they can efficiently synthesize millions of unique neuronal morphologies.

The TNS algorithm is of huge importance for the rapidly growing field of computational neuroscience, which increasingly relies on biologically-realistic models from the single cell level to large-scale neuronal networks. Accurate neuronal morphologies, in particular, lie at the heart of these efforts as they are essential for defining cell types, discerning their functional roles, investigating structural alterations associated with diseased brain states and, identifying what conditions make brain networks sufficiently robust to support the complex cortical processes that are fundamental for a healthy brain.

Therefore, it is essential to accurately reconstruct detailed brain networks in order to replicate the healthy and disease states of the brain.

In a paper published in Cell Reports, a team led by Lida Kanari has applied the Topological Morphology Descriptor (TMD) introduced in Kanari et al. 2018, which reliably categorizes dendritic morphologies, to digitally synthesize dendritic morphologies from all layers and morphological types of the rodent cortex. The advantages of this topology-driven approach are multiple, as the new TNS algorithm, is generalizable to new types of cells, needs little input data and does not require fine tuning because it captures feature correlations.

Enabling the rapid digital reconstruction of entire brain regions from relatively few reference cells

The TNS algorithm driven by the topological architecture of dendrites generates realistic morphologies for a large number of distinct cortical neuronal cell types with realistic morphological and electrical properties. This has enabled the rapid digital reconstruction of entire brain regions from relatively few reference cells, thereby allowing the investigation of links between neuronal morphologies and brain function across different spatio-temporal scales and addressing the challenge of insufficient biological reconstructions.

A multi-stage validation documented in the paper ensures that the synthesized cells reproduce the shapes of reconstructed neurons with respect to three modalities: 1. Their morphological characteristics, 2. The electrical activity of single cells and, 3. The connectivity of the network they form.

Lida Kanari explains that “the findings are already enabling Blue Brain to build biologically detailed reconstructions and simulations of the mouse brain, by computationally reconstructing brain regions for simulations which replicate the anatomical properties of neuronal morphologies and include region specific anatomy. We address one of the fundamental problems for neuroscience—the scarcity of experimental neuronal reconstructions since the topological synthesis requires only a few examples to generate large numbers of unique cells. Using the TNS algorithm, we can efficiently synthesize millions of unique neuronal morphologies (10 million cells in a few hours),” she concludes.

Facilitating medical applications

“Comprehensive neuron models are essential for defining cell types, discerning their functional roles and investigating structural alterations associated with diseased brain states,” affirms Blue Brain Founder and Director, Prof. Henry Markram. “The researchers synthesized cortical networks based on structural alterations of dendrites associated with medical conditions and revealed principles linking branching properties to the structure of large-scale networks.”

“As the TNS algorithm is implemented in an open source software, this will allow the modeling of brain diseases in terms of single cells and networks, as it provides a tool to directly investigate the link between local morphological properties and the connectivity of the neuronal network they form. This approach is of particular interest for medical applications as it enables the investigation of diseases in terms of the emergence of global network pathology from local structural changes in neuron morphologies,” he concludes.

+ Explore further

More information: Lida Kanari et al, Computational synthesis of cortical dendritic morphologies, Cell Reports (2022). DOI: 10.1016/j.celrep.2022.110586

Topological Neuron Synthesis Algorithm: portal.bluebrain.epfl.ch/resou … al-neuron-synthesis/

EPFL and MIT Researchers Discover the ‘Holy Grail’ of Nanowire Production

EPFL researchers have found a way to control and standardize the production of nanowires on silicon surfaces. Credit: Ecole Polytechnique Federale de Lausanne (EPFL)

Nanowires have the potential to revolutionize the technology around us. Measuring just 5-100 nanometers in diameter (a nanometer is a millionth of a millimeter), these tiny, needle-shaped crystalline structures can alter how electricity or light passes through them.

They can emit, concentrate and absorb light and could therefore be used to add optical functionalities to electronic chips. They could, for example, make it possible to generate lasers directly on silicon chips and to integrate single-photon emitters for coding purposes. They could even be applied in solar panels to improve how sunlight is converted into electrical energy.

Up until now, it was impossible to reproduce the process of growing nanowires on silicon semiconductors – there was no way to repeatedly produce homogeneous nanowires in specific positions.

But researchers from EPFL’s Laboratory of Semiconductor Materials, run by Anna Fontcuberta i Morral, together with colleagues from MIT and the IOFFE Institute, have come up with a way of growing nanowire networks in a highly controlled and fully reproducible manner. The key was to understand what happens at the onset of nanowire growth, which goes against currently accepted theories. Their work has been published in Nature Communications.

“We think that this discovery will make it possible to realistically integrate a series of nanowires on silicon substrates,” says Fontcuberta i Morral. “Up to now, these nanowires had to be grown individually, and the process couldn’t be reproduced.”

Two different configurations of the droplet within the opening – hole fully filled and partially filled and bellow illustration of GaAs crystals forming a full ring or a step underneath the large and small gallium droplets. Credit: Ecole Polytechnique Federale de Lausanne (EPFL)

 

Getting the right ratio

The standard process for producing nanowires is to make tiny holes in silicon monoxide and fill them with a nanodrop of liquid gallium. This substance then solidifies when it comes into contact with arsenic. But with this process, the substance tends to harden at the corners of the nanoholes, which means that the angle at which the nanowires will grow can’t be predicted. The search was on for a way to produce homogeneous nanowires and control their position.

Research aimed at controlling the production process has tended to focus on the diameter of the hole, but this approach has not paid off. Now EPFL researchers have shown that by altering the diameter-to-height ratio of the hole, they can perfectly control how the nanowires grow. At the right ratio, the substance will solidify in a ring around the edge of the hole, which prevents the nanowires from growing at a non-perpendicular angle. And the researchers’ process should work for all types of nanowires.

“It’s kind of like growing a plant. They need water and sunlight, but you have to get the quantities right,” says Fontcuberta i Morral.

This new production technique will be a boon for nanowire research, and further samples should soon be developed.

 Explore further: Nanowires have the power to revolutionize solar energy (w/ video)

More information: J. Vukajlovic-Plestina et al. Fundamental aspects to localize self-catalyzed III-V nanowires on silicon, Nature Communications (2019). DOI: 10.1038/s41467-019-08807-9

Journal reference: Nature Communications 

 

A Treasure Trove for Nanotechnology Experts

EPFL and NCCR-MARVEL scientists identified more than 1,000 2-D materials, focusing on the feasibility of exfoliation. Credit: EPFL/G.Pizzi

A team from EPFL and NCCR Marvel has identified more than 1,000 materials with a particularly interesting 2-D structure. Their research, published in Nature Nanotechnlogy, paves the way for groundbreaking technological applications.

2-D materials, which consist of a few layers of atoms, are considered the future of nanotechnology. They offer potential new applications and could be used in small, higher-performance and more energy-efficient devices. Two-dimensional materials were first discovered almost 15 years ago, but only a few dozen of them have been synthesized so far.

Now, thanks to an approach developed by researchers from EPFL’s Theory and Simulation of Materials Laboratory (THEOS) and from NCCR-MARVEL for Computational Design and Discovey of Novel Materials, many more promising 2-D materials may be identified. Their work was recently published in the journal Nature Nanotechnology.

The first 2-D material isolated was graphene, in 2004, earning its discoverers a Nobel Prize in 2010. This marked the start of a whole new era in electronics, as graphene is light, transparent and resilient and, above all, a good conductor of electricity. It paved the way to new applications in such fields as photovoltaics and optoelectronics. “To find other materials with similar properties, we focused on the feasibility of exfoliation,” explains Nicolas Mounet, a researcher in the THEOS lab and lead author of the study.

“But instead of placing adhesive strips on graphite to see if the layers peeled off, like the Nobel Prize winners did, we used a digital method.”

The researchers developed an algorithm to review and carefully analyze the structure of more than 100,000 3-D materials recorded in external databases. From this, they created a database of around 5,600 potential 2-D materials, including more than 1,000 with particularly promising properties. In other words, they’ve created a treasure trove for nanotechnology experts.

To build their database, the researchers used a step-by-step process of elimination. First, they identified all of the materials that are made up of separate layers. “We then studied the chemistry of these materials in greater detail and calculated the energy that would be needed to separate the layers, focusing primarily on materials where interactions between atoms of different layers are weak, something known as Van der Waals bonding,” says Marco Gibertini, a researcher at THEOS and the second author of the study.

Of the 5,600 materials initially identified, the researchers singled out 1,800 structures that could potentially be exfoliated, including 1,036 that looked especially easy to exfoliate. This represents a considerable increase in the number of possible 2-D materials known today. They then selected the 258 most promising materials, categorizing them according to their magnetic, electronic, mechanical, thermal and topological properties.

“Our study demonstrates that digital techniques can really boost discoveries of new materials,” says Nicola Marzari, the director of NCCR-MARVEL and a professor at THEOS. “In the past, chemists had to start from scratch and just keep trying different things, which required hours of lab work and a certain amount of luck. With our approach, we can avoid this long, frustrating process because we have a tool that can single out the materials that are worth studying further, allowing us to conduct more focused research.”

It is also possible to reproduce the researchers’ calculations thanks to their software AiiDA, which describes the calculation process for each material discovered in the form of workflows and stores the full provenance of each stage of the calculation.

“Without AiiDA, it would have been very difficult to combine and process different types of data,” explains Giovanni Pizzi, a senior researcher at THEOS and co-author of the study. “Our workflows are available to the public, so anyone in the world can reproduce our calculations and apply them to any material to find out if it can be exfoliated.

More information: Nicolas Mounet et al, Two-dimensional materials from high-throughput computational exfoliation of experimentally known compounds, Nature Nanotechnology (2018). DOI: 10.1038/s41565-017-0035-5

Provided by Ecole Polytechnique Federale de Lausanne

Explore further:Splitting crystals for 2-D metallic conductivity

Making Cheaper (Perovskite) Solar Cells with 20.2 Percent Efficiency

EPFL scientists have developed a solar-panel material that can cut down on photovoltaic costs while achieving competitive power-conversion efficiency of 20.2%.

Some of the most promising solar cells today use light-harvesting films made from perovskites – a group of materials that share a characteristic molecular structure. However, perovskite-based solar cells use expensive “hole-transporting” materials, whose function is to move the positive charges that are generated when light hits the perovskite film. Publishing in Nature Energy (“A molecularly engineered hole-transporting material for e cient perovskite solar cells”), EPFL scientists have now engineered a considerably cheaper hole-transporting material that costs only a fifth of existing ones while keeping the efficiency of the solar cell above 20%.

This is a 3-D illustration of FDT molecules on a surface of perovskite crystals. (Image: Sven M. Hein / EPFL)

As the quality of perovskite films increases, researchers are seeking other ways of improving the overall performance of solar cells. Inadvertently, this search targets the other key element of a solar panel, the hole-transporting layer, and specifically, the materials that make them up. There are currently only two hole-transporting materials available for perovskite-based solar cells. Both types are quite costly to synthesize, adding to the overall expense of the solar cell.

To address this problem, a team of researchers led by Mohammad Nazeeruddin at EPFL developed a molecularly engineered hole-transporting material, called FDT, that can bring costs down while keeping efficiency up to competitive levels. Tests showed that the efficiency of FDT rose to 20.2% – higher than the other two, more expensive alternatives. And because FDT can be easily modified, it acts as a blueprint for an entire generation of new low-cost hole-transporting materials.

“The best performing perovskite solar cells use hole transporting materials, which are difficult to make and purify, and are prohibitively expensive, costing over €300 per gram preventing market penetration,” says Nazeeruddin. “By comparison, FDT is easy to synthesize and purify, and its cost is estimated to be a fifth of that for existing materials – while matching, and even surpassing their performance.”

Source: Ecole Polytechnique Fédérale de Lausanne

Graphene-perovskite hybrids make new super-detectors: Turning Light into Energy

EPFL scientists have created the first perovskite nanowire-graphene hybrid phototransistors. Even at room temperature, the devices are highly sensitive to light, making them outstanding photodetectors.

The lead-containing perovskite materials can turn light into electricity with high efficiency, which is why they have revolutionized solar cell technologies. On the other hand, graphene is known for its super-strength as well as its excellent electrical conductivity. Combining the two materials, EPFL scientists have created the first ever class of hybrid transistors that turn light into electricity with high sensitivity and at room temperature. The work is published in Small.

The lab of László Forró at EPFL, where the chemical activity is led by Endre Horváth, used its expertise in microengineering to create nanowires of the perovskite methylammonium lead iodide. This highly non-trivial route for the synthesis of nanowires was developed by him in 2014 and called slip-coating method. The advantage of nanowires is their consistency, while their manufacturing can be controlled to modify their architecture and explore different designs.

Making a device by depositing the perovskite nanowires onto graphene has increased the efficiency in converting light to electrical current at room temperature. “Such a device shows almost 750,000 times higher photoresponse compared to detectors made only with perovskite nanowires,” added Massimo Spina who fabricated the miniature photodetectors. Because of this exceptionally high sensitivity, the graphene/perovskite nanowire hybrid device is considered to be a superb candidate for even a single-photon detection.

This work was founded by the Swiss National Science Foundation. The hybrid devices were fabricated in part at EPFL’s Center for Micro/Nanotechnology.

Reference