Connect with us

Science

Where once were black boxes, new LANTERN illuminates

Published

on

Where once were black boxes, new LANTERN illuminates

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.

Described in a new paper published in the Proceedings of the National Academy of Sciences, LANTERN shows the ability to predict the genetic edits needed to create useful differences in three different proteins. One is the spike-shaped protein from the surface of the SARS-CoV-2 virus that causes COVID-19; understanding how changes in the DNA can alter this spike protein might help epidemiologists predict the future of the pandemic. The other two are well-known lab workhorses: the LacI protein from the E. coli bacterium and the green fluorescent protein (GFP) used as a marker in biology experiments. Selecting these three subjects allowed the NIST team to show not only that their tool works, but also that its results are interpretable — an important characteristic for industry, which needs predictive methods that help with understanding of the underlying system.

“We have an approach that is fully interpretable and that also has no loss in predictive power,” said Peter Tonner, a statistician and computational biologist at NIST and LANTERN’s main developer. “There’s a widespread assumption that if you want one of those things you can’t have the other. We’ve shown that sometimes, you can have both.”

Advertisement

The problem the NIST team is tackling might be imagined as interacting with a complex machine that sports a vast control panel filled with thousands of unlabeled switches: The device is a gene, a strand of DNA that encodes a protein; the switches are base pairs on the strand. The switches all affect the device’s output somehow. If your job is to make the machine work differently in a specific way, which switches should you flip?

Because the answer might require changes to multiple base pairs, scientists have to flip some combination of them, measure the result, then choose a new combination and measure again. The number of permutations is daunting.

“The number of potential combinations can be greater than the number of atoms in the universe,” Tonner said. “You could never measure all the possibilities. It’s a ridiculously large number.”

Because of the sheer quantity of data involved, DNNs have been tasked with sorting through a sampling of data and predicting which base pairs need to be flipped. At this, they have proved successful — as long as you don’t ask for an explanation of how they get their answers. They are often described as “black boxes” because their inner workings are inscrutable.

“It is really difficult to understand how DNNs make their predictions,” said NIST physicist David Ross, one of the paper’s co-authors. “And that’s a big problem if you want to use those predictions to engineer something new.”

Advertisement

LANTERN, on the other hand, is explicitly designed to be understandable. Part of its explainability stems from its use of interpretable parameters to represent the data it analyzes. Rather than allowing the number of these parameters to grow extraordinarily large and often inscrutable, as is the case with DNNs, each parameter in LANTERN’s calculations has a purpose that is meant to be intuitive, helping users understand what these parameters mean and how they influence LANTERN’s predictions.

The LANTERN model represents protein mutations using vectors, widely used mathematical tools often portrayed visually as arrows. Each arrow has two properties: Its direction implies the effect of the mutation, while its length represents how strong that effect is. When two proteins have vectors that point in the same direction, LANTERN indicates that the proteins have similar function.

These vectors’ directions often map onto biological mechanisms. For example, LANTERN learned a direction associated with protein folding in all three of the datasets the team studied. (Folding plays a critical role in how a protein functions, so identifying this factor across datasets was an indication that the model functions as intended.) When making predictions, LANTERN just adds these vectors together — a method that users can trace when examining its predictions.

Other labs had already used DNNs to make predictions about what switch-flips would make useful changes to the three subject proteins, so the NIST team decided to pit LANTERN against the DNNs’ results. The new approach was not merely good enough; according to the team, it achieves a new state of the art in predictive accuracy for this type of problem.

“LANTERN equaled or outperformed nearly all alternative approaches with respect to prediction accuracy,” Tonner said. “It outperforms all other approaches in predicting changes to LacI, and it has comparable predictive accuracy for GFP for all except one. For SARS-CoV-2, it has higher predictive accuracy than all alternatives other than one type of DNN, which matched LANTERN’s accuracy but didn’t beat it.”

Advertisement

LANTERN figures out which sets of switches have the largest effect on a given attribute of the protein — its folding stability, for example — and summarizes how the user can tweak that attribute to achieve a desired effect. In a way, LANTERN transmutes the many switches on our machine’s panel into a few simple dials.

“It reduces thousands of switches to maybe five little dials you can turn,” Ross said. “It tells you the first dial will have a big effect, the second will have a different effect but smaller, the third even smaller, and so on. So as an engineer it tells me I can focus on the first and second dial to get the outcome I need. LANTERN lays all this out for me, and it’s incredibly helpful.”

Rajmonda Caceres, a scientist at MIT’s Lincoln Laboratory who is familiar with the method behind LANTERN, said she values the tool’s interpretability.

“There are not a lot of AI methods applied to biology applications where they explicitly design for interpretability,” said Caceres, who is not affiliated with the NIST study. “When biologists see the results, they can see what mutation is contributing to the change in the protein. This level of interpretation allows for more interdisciplinary research, because biologists can understand how the algorithm is learning and they can generate further insights about the biological system under study.”

Tonner said that while he is pleased with the results, LANTERN is not a panacea for AI’s explainability problem. Exploring alternatives to DNNs more widely would benefit the entire effort to create explainable, trustworthy AI, he said.

Advertisement

“In the context of predicting genetic effects on protein function, LANTERN is the first example of something that rivals DNNs in predictive power while still being fully interpretable,” Tonner said. “It provides a specific solution to a specific problem. We hope that it might apply to others, and that this work inspires the development of new interpretable approaches. We don’t want predictive AI to remain a black box.”

Read More

Continue Reading
Advertisement
Click to comment

Leave a Reply

Your email address will not be published.

Science

Global Warming Causes Fewer Tropical Cyclones

Published

on

Global Warming Causes Fewer Tropical Cyclones

Advertisement

But having fewer hurricanes and typhoons does not make them less of a threat. Those that do manage to form are more likely to reach higher intensities as the world continues to heat up with the burning of fossil fuels.


Scientists have been trying for decades to answer the question of how climate change will affect tropical cyclones, given the large-scale death and destruction these storms can cause. Climate models have suggested the number of storms should decline as global temperatures rise, but that had not been confirmed in the historical record. Detailed tropical cyclone data from satellites only go back until about the 1970s, which is not long enough to pick out trends driven by global warming.


The new study worked around those limitations by using what is called a reanalysis: the highest-quality available observations are fed into a weather computer model. “That’s something which gets us close to what the observation would have looked like,” essentially “filling in the gaps,” says study co-author Savin Chand, an atmospheric scientist at Federation University Australia. This gives researchers a reasonably realistic picture of the atmosphere over time, in this case going back to 1850. Chand and his team developed an algorithm that could pick out tropical cyclones in that reanalysis data set, enabling them to look for trends over a 162-year period.

Advertisement


They found the 13 percent global decrease in tropical cyclones over the period of 1900 to 2012, compared with 1850 to 1900 (the latter is widely considered a pre-global-warming reference period). There was an even larger decline of about 23 percent since around 1950, around the time global temperatures started to noticeably rise. The declines vary in different parts of the ocean. For example, the western North Pacific saw 9 percent fewer storms, and the eastern North Pacific saw 18 percent fewer over the 20th and early 21st centuries. And the North Atlantic results indicated a peculiar trend, showing an overall decrease over the past century—but with an uptick in recent decades. That shorter-term increase could be linked to natural climate variations, better detection of storms or a decrease in aerosol pollution (because aerosols have a cooling effect, and tropical cyclones thrive on warm waters).


The study provides crucial ground-truth information for evaluating climate model projections of further future changes in cyclone frequency, says Kimberly Wood, a tropical meteorologist at Mississippi State University, who was not involved with the paper.


Advertisement

Chand and his colleagues link the decrease in tropical storm frequency to changes in atmospheric conditions that constrict convection—the process where warm, moist air surges upward in the atmosphere, which allows tropical cyclones to develop from small weather disturbances that act as the “seeds.” The researchers think those changes are caused by warming-driven shifts in global atmospheric circulation patterns. “It’s a pretty holistic view,” Wood says of the analysis.


But even if there are fewer tropical cyclones overall, a larger proportion of those that do form are expected to reach higher intensities because global warming is also raising sea-surface temperatures and making the atmosphere warmer and moister—the conditions these storms thrive on. “Once a tropical cyclone forms,” Chand says, “there is a lot of fuel in the atmosphere.”

ABOUT THE AUTHOR(S)

author-avatar

    Andrea Thompson, an associate editor at Scientific American, covers sustainability. Follow her on Twitter @AndreaTWeather Credit: Nick Higgins

    Read More

    Advertisement
    Continue Reading

    Science

    The effect of breast cancer screening is declining

    Published

    on

    The effect of breast cancer screening is declining

    Screening for breast cancer has a cost. This is shown by a Danish/Norwegian study that analysed 10,580 breast cancer deaths among Norwegian women aged 50 to 75 years.

    “The beneficial effect of screening is currently declining because the treatment of cancer is improving. Over the last 25 years, the mortality rate for breast cancer has been virtually halved,” says Henrik Støvring, who is behind the study.

    According to the researcher, the problem is that screenings lead to both overdiagnosis and overtreatment, which has a cost both on a human level and in terms of the economy.

    Overdiagnosis and overtreatment

    Advertisement

    When the screening was introduced, the assessment was that around twenty per cent of the deaths from breast cancer among those screened could be averted. While this corresponded to approximately 220 deaths a year in Denmark 25 years ago, today the number has been halved.

    The study shows that in 1996 it was necessary to invite 731 women to avoid a single breast cancer death in Norway, you would have to invite at least 1364 and probably closer to 3500 to achieve the same result in 2016.

    On the other hand, the adverse effects of screening are unchanged.

    “One in five women aged 50-70, who is told they have breast cancer, has received a ‘superfluous’ diagnosis because of screening — without screening, they would never have noticed or felt that they had breast cancer during their lifetime,” says the researcher.

    One in five corresponds to 900 women annually in Denmark. In addition, every year more than 5000 women are told that the screening has given rise to suspicion of breast cancer — a suspicion that later turns out to be incorrect.

    Advertisement

    Peaceful, small nodes — but in who?

    Henrik Støvring notes that the result is not beneficial for the screening programmes. According to him, the Norwegian results can also be transferred to Denmark. Here, women between 50 and 69 are offered a mammogram screening every second year. This is an X-ray examination of the breast, which can show whether the woman has cellular changes that could be breast cancer.

    The Danish screening programme became a national programme offered to all woman in the age group in 2007 — three years after the Norwegians. Approx. 300,000 Danish women are invited to screening for breast cancer every year.

    According to the researcher, the challenge is that we are not currently able to tell the difference between the small cancer tumours that will kill you and those that will not. Some of these small nodes are so peaceful or slow-growing that the woman would die a natural death with undetected breast cancer, if she had not been screened. But once a cancer node has been discovered, it must of course be treated, even though this was not necessary for some of the women — we just do not know who.

    “The women who are invited to screening live longer because all breast cancer patients live longer, and because we have got better drugs, more effective chemotherapy, and because we now have cancer care pathways, which mean the healthcare system reacts faster than it did a decade ago,” says Henrik Støvring.

    Advertisement

    Story Source:

    Materials provided by Aarhus University. Original written by Helle Horskjær Hansen. Note: Content may be edited for style and length.

    Read More

    Advertisement
    Continue Reading

    Science

    Thin-film photovoltaic technology combines efficiency and versatility

    Published

    on

    Thin-film photovoltaic technology combines efficiency and versatility

    Stacking solar cells increases their efficiency. Working with partners in the EU-funded PERCISTAND project, researchers at the Karlsruhe Institute of Technology (KIT) have produced perovskite/CIS tandem solar cells with an efficiency of nearly 25percent- the highest value achieved thus far with this technology. Moreover, this combination of materials is light and versatile, making it possible to envision the use of these tandem solar cells in vehicles, portable equipment, and devices that can be folded or rolled up. The researchers present their results in the journal ACS Energy Letters.

    Perovskite solar cells have made astounding progress over the past decade. Their efficiency is now comparable to that of the long-established silicon solar cells. Perovskites are innovative materials with a special crystal structure. Researchers worldwide are working to get perovskite photovoltaic technology ready for practical applications. The more electricity they generate per unit of surface area, the more attractive solar cells are for consumers

    The efficiency of solar cells can be increased by stacking two or more cells. If each of the stacked solar cells is especially efficient at absorbing light from a different part of the solar spectrum, inherent losses can be reduced and efficiency boosted. The efficiency is a measure of how much of the incident light is converted into electricity. Thanks to their versatility, perovskite solar cells make outstanding components for such tandems. Tandem solar cells using perovskites and silicon have reached a record efficiency level of over 29percent, considerably higher than that of individual cells made of perovskite (25.7percent) or silicon (26.7percent).

    Combining Perovskites with CIS for Mobility and Flexibility

    Advertisement

    Combining perovskites with other materials such as copper-indium-diselenide (CIS) or copper-indium-gallium-diselenide (CIGS) promises further benefits. Such combinations will make it possible to produce light and flexible tandem solar cells that can be installed not only on buildings but also on vehicles and portable equipment. Such solar cells could even be folded or rolled up for storage and extended when needed, for example on blinds or awnings to provide shade and generate electricity at the same time.

    An international team of researchers headed by Dr. Marco A. Ruiz-Preciado and tenure-track professor Ulrich W. Paetzold from the Light Technology Institute (LTI) and the Institute of Microstructure Technology (IMT) at KIT has succeeded in producing perovskite/CIS tandem solar cells with a maximum efficiency of 24.9percent (23.5percent certified). “This is the highest reported efficiency for this technology and the first high efficiency level reached at all with a nearly gallium-free copper-indium diselenide solar cell in a tandem,” says Ruiz-Preciado. Reducing the amount of gallium results in a narrow band gap of approximately one electron volt (eV), which is very close to the ideal value of 0.96eV for the lower solar cell in a tandem.

    CIS Solar Cells with Narrow Band Gap- Perovskite Solar Cells with Low Bromine Content

    The band gap is a material characteristic that determines the part of the solar spectrum that a solar cell can absorb to generate electricity. In a monolithic tandem solar cell, the band gaps must be such that the two cells can produce similar currents to achieve maximum efficiency. If the lower cell’s band gap changes, the upper cell’s band gap has to be adjusted to the change, and vice versa.

    To adjust the band gap for efficient tandem integration, perovskites with high bromine content are usually used. However, this often leads to voltage drops and phase instability. Since the KIT researchers and their partners use CIS solar cells with a narrow band gap at the base of their tandems, they can produce their upper cells using perovskites with low bromine content, which results in cells that are more stable and efficient.

    Advertisement

    “Our study demonstrates the potential of perovskite/CIS tandem solar cells and establishes the foundation for future development to make further improvements in their efficiency,” says Paetzold. “We’ve reached this milestone thanks to the outstanding cooperation in the EU’s PERCISTAND project and, in particular, thanks to our close cooperation with the Netherlands Organisation for Applied Scientific Research.” Important groundwork was done in the CAPITANO project funded by Germany’s Federal Ministry for Economic Affairs and Climate Action (BMWK).

    Story Source:

    Materials provided by Karlsruher Institut für Technologie (KIT). Note: Content may be edited for style and length.

    Read More

    Advertisement
    Continue Reading

    Trending