Artificial Intelligence Edges Closer to the Clinic

TransMED can help predict the outcomes of COVID-19 patients, generating predictions from different kinds of clinical data, including clinical notes, laboratory tests, diagnosis codes and prescribed drugs. The other uniqueness of TransMED lies in its ability to transfer learn from existing diseases to better predict and reason about progression of new and rare diseases. 
Credit: Shannon Colson | Pacific Northwest National Laboratory

The beginning of the COVID-19 pandemic presented a huge challenge to healthcare workers. Doctors struggled to predict how different patients would fare under treatment against the novel SARS-CoV-2 virus. Deciding how to triage medical resources when presented with very little information took a mental and physical toll on caregivers as the pandemic progressed.

To ease this burden, researchers at Pacific Northwest National Laboratory (PNNL), Stanford University, Virginia Tech, and John Snow Labs developed TransMED, a first-of-its-kind artificial intelligence (AI) prediction tool aimed at addressing issues caused by emerging or rare diseases.

“As COVID-19 unfolded over 2020, it brought a number of us together into thinking how and where we could contribute meaningfully,” said chief scientist Sutanay Choudhury. “We decided we could make the most impact if we worked on the problem of predicting patient outcomes.”

“COVID presented a unique challenge,” said Khushbu Agarwal, lead author of the study published in Nature Scientific Reports. “We had very limited patient data for training an AI model that could learn the complex patterns underlying COVID patient trajectories.”

The multi-institutional team developed TransMED to address this challenge, analyzing data from existing diseases to predict outcomes of an emerging disease.

AI tackles the challenge of materials structure prediction

The researchers from Cambridge and Linkoping Universities, have designed a way to predict the structure of materials given its constitutive elements. The results are reported in the journal Science Advances.

The arrangement of atoms in a material determines its properties. The ability to predict this arrangement computationally for different combinations of elements, without having to make the material in the lab, would enable researchers to quickly design and improve materials. This paves the way for advances such as better batteries and photovoltaics.

However, there are many ways that atoms can ‘pack’ into a material: some packings are stable, others are not. Determining the stability of a packing is computationally intensive, and calculating every possible arrangement of atoms to find the best one is not practical. This is a significant bottleneck in materials science.

“This materials structure prediction challenge is similar to the protein folding problem in biology,” said Dr Alpha Lee from Cambridge’s Cavendish Laboratory, who co-led the research. “There are many possible structures that a material can ‘fold’ into. Except the materials science problem is perhaps even more challenging than biology because it considers a much broader set of elements.”

Robots play with play dough

The inner child in many of us feels an overwhelming sense of joy when stumbling across a pile of the fluorescent, rubbery mixture of water, salt, and flour that put goo on the map: play dough. (Even if this happens rarely in adulthood.)

While manipulating play dough is fun and easy for 2-year-olds, the shapeless sludge is hard for robots to handle. Machines have become increasingly reliable with rigid objects, but manipulating soft, deformable objects comes with a laundry list of technical challenges, and most importantly, as with most flexible structures, if you move one part, you’re likely affecting everything else.

Scientists from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and Stanford University recently let robots take their hand at playing with the modeling compound, but not for nostalgia’s sake. Their new system learns directly from visual inputs to let a robot with a two-fingered gripper see, simulate, and shape doughy objects. “RoboCraft” could reliably plan a robot’s behavior to pinch and release play dough to make various letters, including ones it had never seen. With just 10 minutes of data, the two-finger gripper rivaled human counterparts that teleoperated the machine — performing on-par, and at times even better, on the tested tasks.

Where Once Were Black Boxes, NIST’s New LANTERN Illuminates

How do you figure out how to alter a gene so that it makes a usefully different protein? The job might be imagined as interacting with a complex machine (at left) that sports a vast control panel filled with thousands of unlabeled switches, which all affect the device’s output somehow. A new tool called LANTERN figures out which sets of switches — rungs on the gene’s DNA ladder — have the largest effect on a given attribute of the protein. It also summarizes how the user can tweak that attribute to achieve a desired effect, essentially transmuting the many switches on our machine’s panel into another machine (at right) with just a few simple dials.
Credit: B. Hayes/NIST

Researchers at the National Institute of Standards and Technology (NIST) have developed a new statistical tool that they have used to predict protein function. Not only could it help with the difficult job of altering proteins in practically useful ways, but it also works by methods that are fully interpretable — an advantage over the conventional artificial intelligence (AI) that has aided with protein engineering in the past.

The new tool, called LANTERN, could prove useful in work ranging from producing biofuels to improving crops to developing new disease treatments. Proteins, as building blocks of biology, are a key element in all these tasks. But while it is comparatively easy to make changes to the strand of DNA that serves as the blueprint for a given protein, it remains challenging to determine which specific base pairs — rungs on the DNA ladder — are the keys to producing a desired effect. Finding these keys has been the purview of AI built of deep neural networks (DNNs), which, though effective, are notoriously opaque to human understanding.

AI reveals unsuspected math underlying search for exoplanets

Artist’s concept of a sun-like star (left) and a rocky planet about 60% larger than Earth in orbit in the star’s habitable zone. Gravitational microlensing has the ability to detect such planetary systems and determine the masses and orbital distances, even though the planet itself is too dim to be seen. 
Image credit: NASA Ames/JPL-Caltech/T. Pyle

Artificial intelligence (AI) algorithms trained on real astronomical observations now outperform astronomers in sifting through massive amounts of data to find new exploding stars, identify new types of galaxies and detect the mergers of massive stars, accelerating the rate of new discovery in the world’s oldest science.

But AI, also called machine learning, can reveal something deeper, University of California, Berkeley, astronomers found: unsuspected connections hidden in the complex mathematics arising from general relativity — in particular, how that theory is applied to finding new planets around other stars.

In a paper appearing this week in the journal Nature Astronomy, the researchers describe how an AI algorithm developed to more quickly detect exoplanets when such planetary systems pass in front of a background star and briefly brighten it — a process called gravitational microlensing — revealed that the decades-old theories now used to explain these observations are woefully incomplete.

Artificial intelligence predicts patients’ race from their medical images

Researchers demonstrated that medical AI systems can easily learn to recognize racial identity in medical images, and that this capability is extremely difficult to isolate or mitigate.
 Credit: Massachusetts Institute of Technology

The miseducation of algorithms is a critical problem; when artificial intelligence mirrors unconscious thoughts, racism, and biases of the humans who generated these algorithms, it can lead to serious harm. Computer programs, for example, have wrongly flagged Black defendants as twice as likely to reoffend as someone who’s white. When an AI used cost as a proxy for health needs, it falsely named Black patients as healthier than equally sick white ones, as less money was spent on them. Even AI used to write a play relied on using harmful stereotypes for casting.

Removing sensitive features from the data seems like a viable tweak. But what happens when it’s not enough?