Highlights from Latest Articles

New for April 18, 2018

Agricultural Sciences; Computer Sciences

Using machine learning to examine plant stresses

Plant stresses are typically classified by trained experts who manually identify visual symptoms, a process that is time-consuming, error-prone, and subjective. Sambuddha Ghosal et al. developed a machine learning approach to automate the process of identifying, classifying, and quantifying stresses in soybean. The authors trained the deep learning model on approximately 25,000 images of stressed and healthy soybean leaflets, and the trained model rapidly and accurately identified and classified several abiotic stresses, such as herbicide injury and nutrient deficiency, and biotic stresses, such as fungal and bacterial diseases, from images of soybean leaflets. To determine which visual symptoms the model used to make its predictions, the authors extracted high-resolution feature maps learned by the model in an unsupervised manner and correlated the maps with the visual symptoms of each stress identified by human experts. The authors found a high level of agreement between the machine and human annotations. The model also accurately identified stresses in other plant species and under different lighting and imaging conditions, suggesting that the approach could be widely applicable for real-time, high-throughput phenotyping of plant stresses. According to the authors, the deep learning approach could have a variety of applications in plant research, plant breeding, and crop production and management. — S.R.

"An explainable deep machine vision framework for plant stress phenotyping," by Sambuddha Ghosal, et al.
Article

---

Chemistry; Microbiology

Electroactive bacteria as a catalyst regulator

Metabolic engineering harnesses biological processes to produce molecules of interest, including pharmaceuticals and fuels. To expand the types of reactions amenable to metabolic engineering, Gang Fan et al. used the bacterium Shewanella oneidensis to control the activity of copper-based catalysts in a radical polymerization reaction. Although the contents of cells released during lysis could effect the polymerization reaction, the authors found that polymerization rates in cases of lysis were low and that dead intact cells resulted in no reaction, suggesting that the metabolic activity of live cells was responsible for powering the polymerization reaction. The polymer products exhibited a well-defined molecular weight, and using other bacteria such as Escherichia coli produced no significant polymerization. The rate of polymerization was dependent on the carbon source for S. oneidensis and specific electron transport proteins. According to the authors, the factors that affect polymer production, such as carbon substrates, catalyst structure, and electron transport proteins, demonstrate the tunable nature of the reaction and suggest that bacterial metabolism can be coupled to redox-active metal catalysts to drive chemical reactions beyond polymerization. — P.G.

"Shewanella oneidensis as a living electrode for controlled radical polymerization," by Gang Fan, Christopher M. Dundas, Austin J. Graham, Nathaniel A. Lynd, and Benjamin K. Keitz
Article

---

Ecology

Compounds tied to malaria infection increase human attractiveness to mosquitoes

To locate hosts, female Anopheles mosquitoes use volatile compounds released from human skin that contribute to body odor. Previous studies have found that Plasmodium infection increases human attractiveness to mosquitoes, but the underlying skin chemistry remained unexplored. Ailie Robinson et al. report that in a lab assay Anopheles gambiae mosquitoes were more attracted to the foot odor of children who harbored Plasmodium than to the foot odor of the same children collected 21 days after administration of an antimalarial treatment; foot odors were collected on socks from 5- to 12-year-old children in western Kenya. In contrast, the mosquitoes did not distinguish between foot odors of parasite-free children collected at the same pair of time points. Chemical analysis revealed that increased amounts of the skin aldehydes heptanal, octanal, and nonanal in infected children were positively linked with parasite density. Mosquito behavioral analysis under controlled conditions suggested that elevated heptanal levels in Plasmodium-infected children might heighten their attractiveness to Anopheles coluzzii mosquitoes. Adding heptanal to a synthetic mosquito lure that mimics healthy human odor failed to increase its attractiveness. However, adding a blend of Plasmodium infection-associated compounds, including heptanal, to this lure, in an attempt to mimic the odor cues of an infected person, proved attractive to An. coluzzii. According to the authors, the findings could illuminate aspects of malaria transmission, help improve the design of mosquito lures, and lead to the development of novel diagnostic biomarkers for malaria. — P.N.

"Plasmodium-associated changes in human odor attract mosquitoes," by Ailie Robinson, et al.
Article

---

Engineering

Photovoltaics, thermal stability, and power conversion

Flexible photovoltaic devices offer a potential solution for supplying electrical power to wearable sensors and electronic devices. However, current flexible organic photovoltaics (OPVs) lack the thermal stability and power conversion requirements to enable the optimal performance of such devices. Xiaomin Xu et al. developed an ultraflexible OPV with increased power conversion efficiency (PCE) and thermal stability. The 3-μm-thick device consists of a robust polymer microstructure and an ultrathin plastic substrate and barrier coating that enable high thermal stability. The authors conducted stability and lifetime testing of the OPV in air chambers in which the OPV was exposed to different temperatures. During tests in which the OPV cells were kept at air temperatures of 120 °C and 130 °C, the authors observed decreases of initial PCE of 7.4% and 13.8%, respectively; in comparison, another OPV polymer design exhibited a 37.6% decrease in initial PCE at 130 °C. Moreover, the ultrathin OPV maintained 80% of the initial PCE at over 500 hours of continuous thermal stress at 85 °C. The findings carry implications for power generation for common electronic devices, according to the authors. — C.S.

"Thermally stable, highly efficient, ultraflexible organic photovoltaics," by Xiaomin Xu, et al.
Article

---

Genetics; Psychological and Cognitive Sciences

Genetic variation tied to consonant use across populations

Variations in READ1, a genetic sequence that regulates expression of the gene DCDC2, are tied to dyslexia and thought to influence normal speech variation, reading performance, and auditory processing of basic word sounds. Mellissa DeMille et al. compared READ1 sequences in nonhuman primates, Neanderthals, Denisovans, and modern humans, and found that the regulatory sequence acquired changes specific to modern humans between 550,000 and 4 million years ago. Comparison of READ1 variants across 43 modern human populations from five continents with the numbers of consonants and vowels prevalent in those populations revealed that the number of consonants—but not vowels—was positively correlated with the frequency of RU1-1, a READ1 variant; the human brain uses distinct strategies to process and encode vowels and consonants. The correlation was unaffected by the populations' geographic proximity, genetic relatedness, or linguistic relatedness. Hence, the authors suggest, differences in READ1 sequence, and resultant differences in DCDC2 expression, may account for differences in word-sound processing across populations. Specifically, the low prevalence of RU1-1 in some populations may be associated with diminished ability to distinguish between consonants, resulting in progressive shrinkage of the populations' consonant inventory. According to the authors, subtle but discernible genetic variations, magnified by cultural processes, might account for differences in consonant use across populations. — P.N.

"Worldwide distribution of the DCDC2 READ1 regulatory element and its relationship with phoneme variation across languages,"by Mellissa M. C. DeMille, et al.
Article

---

Some of the highlights have previously appeared on the PNAS media tipsheet. The articles in PNAS report original research by independent authors and do not necessarily represent the view of the National Academy of Sciences.

[04/18]