System Bits: Aug. 14

AI for less toxic cancer treatment; superconductors with a twist; nanoribbons.


Machine-learning system determines the fewest, smallest doses that could still shrink brain tumors
In an effort to improve the quality of life for patients by reducing toxic chemotherapy and radiotherapy dosing for glioblastoma, the most aggressive form of brain cancer, MIT researchers are employing novel machine-learning techniques.

According to the team, glioblastoma is a malignant tumor that appears in the brain or spinal cord, with a prognosis for adults of no more than five years. Patients must endure a combination of radiation therapy and multiple drugs taken every month whereby medical professionals generally administer maximum safe drug doses to shrink the tumor as much as possible but these strong pharmaceuticals still cause debilitating side effects in patients.

MIT researchers aim to improve the quality of life for patients suffering from glioblastoma, the most aggressive form of brain cancer, with a machine-learning model that makes chemotherapy and radiotherapy dosing regimens less toxic but still as effective as human-designed regimens.

Source: MIT

In a paper being presented at the 2018 Machine Learning for Healthcare conference at Stanford University, MIT Media Lab researchers detail a model that could make dosing regimens less toxic but still effective.

Powered by a “self-learning” machine-learning technique, the model looks at treatment regimens currently in use, and iteratively adjusts the doses, eventually finding an optimal treatment plan with the lowest possible potency and frequency of doses that should still reduce tumor sizes to a degree comparable to that of traditional regimens.

In simulated trials of 50 patients, the machine-learning model designed treatment cycles that reduced the potency to a quarter or half of nearly all the doses while maintaining the same tumor-shrinking potential. Many times, it skipped doses altogether, scheduling administrations only twice a year instead of monthly.

The researchers’ model uses a technique called reinforced learning (RL), a method inspired by behavioral psychology, in which a model learns to favor certain behavior that leads to a desired outcome.

The team pointed out that this approach was used to train the computer program DeepMind that in 2016 made headlines for beating one of the world’s best human players in the game “Go.” It’s also used to train driverless cars in maneuvers, such as merging into traffic or parking, where the vehicle will practice over and over, adjusting its course, until it gets it right.

The researchers prove this model offers a major improvement over the conventional “eye-balling” method of administering doses, observing how patients respond, and adjusting accordingly since humans don’t have the in-depth perception that a machine looking at tons of data has, so the human process is slow, tedious, and inexact. Here, you’re just letting a computer look for patterns in the data, which would take forever for a human to sift through, and use those patterns to find optimal doses.

Surprising distortions found in high-temperature superconductors
According to Rice University researchers, there’s a literal disturbance in the force that alters what physicists have long thought of as a characteristic of superconductivity, and which may give researchers better ability to design materials with novel and predictable properties by being able to manipulate the point of optimum doping.

Rice physicists Pengcheng Dai and Andriy Nevidomskyy and their colleagues used simulations and neutron scattering experiments that show the atomic structure of materials to reveal tiny distortions of the crystal lattice in a so-called iron pnictide compound of sodium, iron, nickel and arsenic.

These local distortions were observed among the otherwise symmetrical atomic order in the material at ultracold temperatures near the point of optimal superconductivity. They indicate researchers may have some wiggle room as they work to increase the temperature at which iron pnictides become superconductors.

Rice University researchers used experiments and simulations to discovery small distortions in the lattice of an iron pnictide that becomes superconductive at ultracold temperatures. They suspect these distortions introduce pockets of superconductivity in the material above temperatures at which it becomes entirely superconductive.
Source: Rice University

Dai and Nevidomskyy, both members of the Rice Center for Quantum Materials (RCQM), are interested in the fundamental processes that give rise to novel collective phenomena like superconductivity, which allows materials to transmit electrical current with no resistance.

Scientists originally found superconductivity at ultracold temperatures that let atoms cooperate in ways that aren’t possible at room temperature. Even known “high-temperature” superconductors top out at 134 Kelvin at ambient pressure, equivalent to minus 218 degrees Fahrenheit, so if there’s any hope for widespread practical use of superconductivity, scientists have to find loopholes in the basic physics of how atoms and their constituents behave under a variety of conditions.

That is what the Rice researchers have done with the iron pnictide, an “unconventional superconductor” of sodium, iron and arsenic, especially when doped with nickel.

These single crystals of nickel-doped compounds of sodium, iron and arsenic are like those used by Rice University researchers in experiments to determine the material’s superconductive properties at ultracold temperatures. They used simulations and precise neutron scattering experiments to show the presence of tiny lattice distortions near the optimal superconductivity of an iron pnictide compound.
Source: Rice University

The key to the material’s superconductivity seems to lie within a subtle property that is unique to iron pnictides: a structural transition in its crystal lattice, the ordered arrangement of its atoms, from tetragonal to orthorhombic. In a tetragonal crystal, the atoms are arranged like cubes that have been stretched in one direction. An orthorhombic structure is shaped like a brick.

The researchers believe that being able to manipulate that point of optimum doping may give them better ability to design materials with novel and predictable properties.

Tying electrons down with nanoribbons
Scientists are experimenting with narrow strips of graphene, called nanoribbons, in hopes of making cool new electronic devices, but University of California, Berkeley scientists have discovered another possible role for them: as nanoscale electron traps with potential applications in quantum computers.

Graphene, a sheet of carbon atoms arranged in a rigid, honeycomb lattice resembling chicken wire, has interesting electronic properties of its own. But when scientists cut off a strip less than about 5 nanometers in width – less than one ten-thousandth the width of a human hair – the graphene nanoribbon takes on new quantum properties, making it a potential alternative to silicon semiconductors.
UC Berkeley theoretician Steven Louie, a professor of physics, predicted last year that joining two different types of nanoribbons could yield a unique material, one that immobilizes single electrons at the junction between ribbon segments.

Scanning tunneling microscope image of a topological nanoribbon superlattice. Electrons are trapped at the interfaces between wide ribbon segments (which are topologically non-trivial) and narrow ribbon segments (which are topologically trivial). The wide segments are 9 carbon atoms across (1.65 nanometers) while the narrow segments are only 7 carbon atoms across (1.40 nanometers).
Source: UC Berkeley

In order to accomplish this, however, the electron “topology” of the two nanoribbon pieces must be different. Topology here refers to the shape that propagating electron states adopt as they move quantum mechanically through a nanoribbon, a subtle property that had been ignored in graphene nanoribbons until Louie’s prediction.

Two of Louie’s colleagues, chemist Felix Fischer and physicist Michael Crommie, became excited by his idea and the potential applications of trapping electrons in nanoribbons and teamed up to test the prediction. Together they were able to experimentally demonstrate that junctions of nanoribbons having the proper topology are occupied by individual localized electrons.

A nanoribbon made according to Louie’s recipe with alternating ribbon strips of different widths, forming a nanoribbon superlattice, produces a conga line of electrons that interact quantum mechanically. Depending on the strips’ distance apart, the new hybrid nanoribbon is either a metal, a semiconductor or a chain of qubits, the basic elements of a quantum computer.

“This gives us a new way to control the electronic and magnetic properties of graphene nanoribbons,” said Crommie, a UC Berkeley professor of physics. “We spent years changing the properties of nanoribbons using more conventional methods, but playing with their topology gives us a powerful new way to modify the fundamental properties of nanoribbons that we never suspected existed until now.”
Using the mathematics of topology Louie’s theory implies that nanoribbons are topological insulators: unusual materials that are insulators, that is, non-conducting in the interior, but metallic conductors along their surface. The 2016 Nobel Prize in Physics was awarded to three scientists who first used the mathematical principles of topology to explain strange, quantum states of matter, now classified as topological materials.

Leave a Reply