Bacteria are known for breaking down lactose to make yogurt and sugar to make beer. Now researchers led by Northwestern University and LanzaTech have harnessed bacteria to break down waste carbon dioxide (CO₂) to make valuable industrial chemicals.

In a new pilot study, the researchers selected, engineered and optimized a bacteria strain and then successfully demonstrated its ability to convert CO₂ into acetone and isopropanol (IPA).

Not only does this new gas fermentation process remove greenhouse gases from the atmosphere, it also avoids using fossil fuels, which are typically needed to generate acetone and IPA. After performing life-cycle analysis, the team found the carbon-negative platform could reduce greenhouse gas emissions by 160% as compared to conventional processes, if widely adopted.

The study will be published on Monday (Feb. 21) in the journal Nature Biotechnology.

“The accelerating climate crisis, combined with rapid population growth, pose some of the most urgent challenges to humankind, all linked to the unabated release and accumulation of CO₂ across the entire biosphere,” said Northwestern’s Michael Jewett, co-senior author of the study. “By harnessing our capacity to partner with biology to make what is needed, where and when it is needed, on a sustainable and renewable basis, we can begin to take advantage of the available CO₂ to transform the bioeconomy.”

Jewett is the Walter P. Murphy Professor of Chemical and Biological Engineering at Northwestern’s McCormick School of Engineering and director of the Center for Synthetic Biology. He co-led the study with Michael Koepke and Ching Leang, both researchers at LanzaTech.

Necessary industrial bulk and platform chemicals, acetone and IPA are found nearly everywhere, with a combined global market topping $10 billion. Widely used as a disinfectant and antiseptic, IPA is the basis for one of the two World Health Organization-recommended sanitizer formulas, which are highly effective in killing the SARS-CoV-2 virus. And acetone is a solvent for many plastics and synthetic fibers, thinning polyester resin, cleaning tools and nail polish remover.

While these chemicals are incredibly useful, they are generated from fossil resources, leading to climate-warming CO₂ emissions.

To manufacture these chemicals more sustainably, the researchers developed a new gas fermentation process. They started with Clostridium autoethanogenum, an anaerobic bacterium engineered at LanzaTech. Then, the researchers used synthetic biology tools to reprogram the bacterium to ferment CO₂ to make acetone and IPA.

“These innovations, led by cell-free strategies that guided both strain engineering and optimization of pathway enzymes, accelerated time to production by more than a year,” Jewett said.

The Northwestern and LanzaTech teams believe the developed strains and fermentation process will translate to industrial scale. The approach also could potentially be applied to create streamlined processes for generating other valuable chemicals.

“This discovery is a major step forward in avoiding a climate catastrophe,” said Jennifer Holmgren, LanzaTech CEO. “Today, most of our commodity chemicals are derived exclusively from new fossil resources such as oil, natural gas or coal. Acetone and IPA are two examples with a combined global market of $10 billion. The acetone and IPA pathways developed will accelerate the development of other new products by closing the carbon cycle for their use in multiple industries.”

Jewett is a member of the Chemistry of Life Processes Institute, Simpson Querrey Institute for BioNanotechnology and the Robert H. Lurie Comprehensive Cancer Center of Northwestern University.

Story Source:

Materials provided by Northwestern University. Original written by Amanda Morris. Note: Content may be edited for style and length.

“We’re now moving beyond taking isolated snapshots of specific regions of exoplanet atmospheres, to study them as the 3D systems they truly are,” says Thomas Mikal-Evans, who led the study as a postdoc in MIT’s Kavli Institute for Astrophysics and Space Research.

The planet at the center of the new study, which appears in Nature Astronomy, is WASP-121b, a massive gas giant nearly twice the size of Jupiter. The planet is an ultrahot Jupiter and was discovered in 2015 orbiting a star about 850 light years from Earth. WASP-121b has one of the shortest orbits detected to date, circling its star in just 30 hours. It is also tidally locked, such that its star-facing “day” side is permanently roasting, while its “night” side is turned forever toward space.

“Hot Jupiters are famous for having very bright day sides, but the night side is a different beast. WASP-121b’s night side is about 10 times fainter than its day side,” says Tansu Daylan, a TESS postdoc at MIT who co-authored the study.

Astronomers had previously detected water vapor and studied how the atmospheric temperature changes with altitude on the planet’s day side.

The new study captures a much more detailed picture. The researchers were able to map the dramatic temperature changes from the day to the night side, and to see how these temperatures change with altitude. They also tracked the presence of water through the atmosphere to show, for the first time, how water circulates between a planet’s day and night sides.

While on Earth, water cycles by first evaporating, then condensing into clouds, then raining out, on WASP-121b, the water cycle is far more intense: On the day side, the atoms that make up water are ripped apart at temperatures over 3,000 Kelvin. These atoms are blown around to the night side, where colder temperatures allow hydrogen and oxygen atoms to recombine into water molecules, which then blow back to the day side, where the cycle starts again.

The team calculates that the planet’s water cycle is sustained by winds that whip the atoms around the planet at speeds of up to 5 kilometers per second, or more than 11,000 miles per hour.

It also appears that water isn’t alone in circulating around the planet. The astronomers found that the night side is cold enough to host exotic clouds of iron and corundum — a mineral that makes up rubies and sapphires. These clouds, like water vapor, may whip around to the day side, where high temperatures vaporize the metals into gas form. On the way, exotic rain might be produced, such as liquid gems from the corundum clouds.

“With this observation, we’re really getting a global view of an exoplanet’s meteorology,” Mikal-Evans says.

The study’s co-authors include collaborators from MIT, Johns Hopkins University, Caltech, and other institutions.

Day and night

The team observed WASP-121b using a spectroscopic camera aboard NASA’s Hubble Space Telescope. The instrument observes the light from a planet and its star, and breaks that light down into its constituent wavelengths, the intensities of which give astronomers clues to an atmosphere’s temperature and composition.

Through spectroscopic studies, scientists have observed atmospheric details on the day sides of many exoplanets. But doing the same for the night side is far trickier, as it requires watching for tiny changes in the planet’s entire spectrum as it circles its star.

For the new study, the team observed WASP-121b throughout two full orbits — one in 2018, and the other in 2019. For both observations, the researchers looked through the light data for a specific line, or spectral feature, that indicated the presence of water vapor.

“We saw this water feature and mapped how it changed at different parts of the planet’s orbit,” Mikal-Evans says. “That encodes information about what the temperature of the planet’s atmosphere is doing as a function of altitude.”

The changing water feature helped the team map the temperature profile of both the day and night side. They found the day side ranges from 2,500 Kelvin at its deepest observable layer, to 3,500 Kelvin in its topmost layers. The night side ranged from 1,800 Kelvin at its deepest layer, to 1,500 Kelvin in its upper atmosphere. Interestingly, temperature profiles appeared to flip-flop, rising with altitude on the day side — a “thermal inversion,” in meteorological terms — and dropping with altitude on the night side.

The researchers then passed the temperature maps through various models to identify chemicals that are likely to exist in the planet’s atmosphere, given specific altitudes and temperatures. This modeling revealed the potential for metal clouds, such as iron, corundum, and titanium on the night side.

From their temperature mapping, the team also observed that the planet’s hottest region is shifted to the east of the “substellar” region directly below the star. They deduced that this shift is due to extreme winds.

“The gas gets heated up at the substellar point but is getting blown eastward before it can reradiate to space,” Mikal-Evans explains.

From the size of the shift, the team estimates that the wind speeds clock in at around 5 kilometers per second.

“These winds are much faster than our jet stream, and can probably move clouds across the entire planet in about 20 hours,” says Daylan, who led previous work on the planet using NASA’s MIT-led mission, TESS.

The astronomers have reserved time on the James Webb Space Telescope to observe WASP-121b later this year, and hope to map changes in not just water vapor but also carbon monoxide, which scientists suspect should reside in the atmosphere.

“That would be the first time we could measure a carbon-bearing molecule in this planet’s atmosphere,” Mikal-Evans says. “The amount of carbon and oxygen in the atmosphere provides clues on where these kinds of planet form.”

This research was supported in part by NASA through a grant from the Space Telescope Science Institute.

Video: https://youtu.be/XnMXf-uASzo

For instance, if a dataset contains mostly images of white men, then a facial-recognition model trained with this data may be less accurate for women or people with different skin tones.

A group of researchers at MIT, in collaboration with researchers at Harvard University and Fujitsu, Ltd., sought to understand when and how a machine-learning model is capable of overcoming this kind of dataset bias. They used an approach from neuroscience to study how training data affects whether an artificial neural network can learn to recognize objects it has not seen before. A neural network is a machine-learning model that mimics the human brain in the way it contains layers of interconnected nodes, or “neurons,” that process data.

The new results show that diversity in training data has a major influence on whether a neural network is able to overcome bias, but at the same time dataset diversity can degrade the network’s performance. They also show that how a neural network is trained, and the specific types of neurons that emerge during the training process, can play a major role in whether it is able to overcome a biased dataset.

“A neural network can overcome dataset bias, which is encouraging. But the main takeaway here is that we need to take into account data diversity. We need to stop thinking that if you just collect a ton of raw data, that is going to get you somewhere. We need to be very careful about how we design datasets in the first place,” says Xavier Boix, a research scientist in the Department of Brain and Cognitive Sciences (BCS) and the Center for Brains, Minds, and Machines (CBMM), and senior author of the paper.

Co-authors include former graduate students Spandan Madan, a corresponding author who is currently pursuing a PhD at Harvard, Timothy Henry, Jamell Dozier, Helen Ho, and Nishchal Bhandari; Tomotake Sasaki, a former visiting scientist now a researcher at Fujitsu; Frédo Durand, a professor of electrical engineering and computer science and a member of the Computer Science and Artificial Intelligence Laboratory; and Hanspeter Pfister, the An Wang Professor of Computer Science at the Harvard School of Enginering and Applied Sciences. The research appears today in Nature Machine Intelligence.

Thinking like a neuroscientist

Boix and his colleagues approached the problem of dataset bias by thinking like neuroscientists. In neuroscience, Boix explains, it is common to use controlled datasets in experiments, meaning a dataset in which the researchers know as much as possible about the information it contains.

The team built datasets that contained images of different objects in varied poses, and carefully controlled the combinations so some datasets had more diversity than others. In this case, a dataset had less diversity if it contains more images that show objects from only one viewpoint. A more diverse dataset had more images showing objects from multiple viewpoints. Each dataset contained the same number of images.

The researchers used these carefully constructed datasets to train a neural network for image classification, and then studied how well it was able to identify objects from viewpoints the network did not see during training (known as an out-of-distribution combination).

For example, if researchers are training a model to classify cars in images, they want the model to learn what different cars look like. But if every Ford Thunderbird in the training dataset is shown from the front, when the trained model is given an image of a Ford Thunderbird shot from the side, it may misclassify it, even if it was trained on millions of car photos.

The researchers found that if the dataset is more diverse — if more images show objects from different viewpoints — the network is better able to generalize to new images or viewpoints. Data diversity is key to overcoming bias, Boix says.

“But it is not like more data diversity is always better; there is a tension here. When the neural network gets better at recognizing new things it hasn’t seen, then it will become harder for it to recognize things it has already seen,” he says.

Testing training methods

The researchers also studied methods for training the neural network.

In machine learning, it is common to train a network to perform multiple tasks at the same time. The idea is that if a relationship exists between the tasks, the network will learn to perform each one better if it learns them together.

But the researchers found the opposite to be true — a model trained separately for each task was able to overcome bias far better than a model trained for both tasks together.

“The results were really striking. In fact, the first time we did this experiment, we thought it was a bug. It took us several weeks to realize it was a real result because it was so unexpected,” he says.

They dove deeper inside the neural networks to understand why this occurs.

They found that neuron specialization seems to play a major role. When the neural network is trained to recognize objects in images, it appears that two types of neurons emerge — one that specializes in recognizing the object category and another that specializes in recognizing the viewpoint.

When the network is trained to perform tasks separately, those specialized neurons are more prominent, Boix explains. But if a network is trained to do both tasks simultaneously, some neurons become diluted and don’t specialize for one task. These unspecialized neurons are more likely to get confused, he says.

“But the next question now is, how did these neurons get there? You train the neural network and they emerge from the learning process. No one told the network to include these types of neurons in its architecture. That is the fascinating thing,” he says.

That is one area the researchers hope to explore with future work. They want to see if they can force a neural network to develop neurons with this specialization. They also want to apply their approach to more complex tasks, such as objects with complicated textures or varied illuminations.

Boix is encouraged that a neural network can learn to overcome bias, and he is hopeful their work can inspire others to be more thoughtful about the datasets they are using in AI applications.

This work was supported, in part, by the National Science Foundation, a Google Faculty Research Award, the Toyota Research Institute, the Center for Brains, Minds, and Machines, Fujitsu Laboratories Ltd., and the MIT-Sensetime Alliance on Artificial Intelligence.