Although great progress has been made for Machine Reading Comprehension (MRC) in English, scaling out to a large number of languages remains a huge challenge due to the lack of large amounts of annotated training data in non-English languages. To address this challenge, some recent efforts of cross-lingual MRC employ machine translation to transfer knowledge from English to other languages, through either explicit alignment or implicit attention. For effective knowledge transition, it is beneficial to leverage both semantic and syntactic information. However, the existing methods fail to explicitly incorporate syntax information in model learning. Consequently, the models are not robust to errors in alignment and noises in attention. In this work, we propose a novel approach, named GraFusionMRC, which jointly models the cross-lingual alignment information and the mono-lingual syntax information using a graph. We develop a series of algorithms, including graph construction, learning, and pre-training. The experiments on two benchmark datasets for cross-lingual MRC show that our approach outperforms all strong baselines, which verifies the effectiveness of syntax information for cross-lingual MRC.

A Graph Fusion Approach for Cross-Lingual Machine Reading Comprehension

Although multilingual pretrained models (mPLMs) enabled support of various natural language processing in diverse languages, its limited coverage of 100+ languages lets 6500+ languages remain ‘unseen’. One common approach for an unseen language is specializing the model for it as target, by performing additional masked language modeling (MLM) with the target language corpus. However, we argue that, due to the discrepancy from multilingual MLM pretraining, a naive specialization as such can be suboptimal. Specifically, we pose three discrepancies to overcome. Script and linguistic discrepancy of the target language from the related seen languages, hinder a positive transfer, for which we propose to maximize representation similarity, unlike existing approaches maximizing overlaps. In addition, label space for MLM prediction can vary across languages, for which we propose to reinitialize top layers for a more effective adaptation. Experiments over four different language families and three tasks shows that our method improves the task performance of unseen languages with statistical significance, while previous approach fails to.

Script, Language, and Labels: Overcoming Three Discrepancies for Low-Resource Language Specialization

Magnetic skyrmions are topological spin textures with great potential for spintronics applications (1). The appearance of skyrmions requires a careful balance between the magnetic anisotropy and the Dzyaloshinskii-Moriya interaction, and being able to tune these values with a voltage is an important tool for material optimisation for skyrmion devices.

In this work we have explored generation of both volatile and non-volatile skyrmions in MgO/Mn2CoAl/Pd thin films using different ionic liquid gating voltage sequences and magneto-optical Kerr effect (MOKE) microscopy. With a negative gate voltage of -2.5 V, we are able to generate non-volatile skyrmions ~ 1 μm diameter in films with perpendicular magnetic anisotropy Keff ~ 1 x 104 Jm-3. The same result is achieved in higher Keff films by ‘training’ or repeatedly cycling the gate voltage, achieving a giant voltage tunability of magnetic anisotropy of 109.8 mT V-1. Interestingly, volatile skyrmions can also be generated, by first applying a large negative ‘trigger’ voltage, and then skyrmions appear at a positive gate voltage. Using X-ray photoelectron spectroscopy, we explain the generation of skyrmions in terms of magneto-ionic and electrostatic effects (2). Our results show the potential of ionic liquid gating for achieving large anisotropy changes to reversibly or irreversibly generate skyrmions from materials with a range of starting anisotropy values.<br>

**References:**<br>
(1) X. Zhang, Y. Zhou, K. M. Song, T.-E. Park, J. Xia, M. Ezawa, X. Liu, W. Zhao, G. Zhao, and S. Woo, J. Phys. Cond. Mat. 32, 143001 (2020).<br>
(2) Y. Zhang, G. Dubuis, C. Doyle, T. Butler, and S. Granville, Phys. Rev. Appl. 16, 014030 (2021).<br>

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/21c77abfee9e932ada1d2b3ce6ec1e37.png)


Magneto ionic and electrostatic generation of non

Rice (Oryza sativa L.) is one of the important cereal grains and a staple food for more than 50% of the world’s population. Increases in global temperatures have a negative impact on crop productivity. Plants undergo a variety of physiological and biochemical changes as a result of temperature stress in order to adapt to higher temperatures. In this study, we have used 204 accessions from the Bengal and Assam Aus Panel (BAAP). A total of eight replicate plant per accession were grown. All plants were grown in the greenhouse until maturity; thereafter, five replicate plants were randomly selected, their panicles at anthesis stage marked and were transferred to large growth room at 38°C for 24 hours to assess high-temperature treatments. Three replicates plants were not heat treated. Fertility was assessed as the proportion of flowers filled. Heat treatment significantly (P&lt;0.0001) reduced fertility by 17.4% on average and there was a significant (P&lt;0.0001) differences among accessions. In addition to that, there was a significant (P&lt;0.001) interaction between accession and treatment. Heat tolerant cultivars have been identified including the aus reference cultivar DJ123. Genome Wide Association Studies (GWAS) using 2 million SNPs will be used to identify QTLs and candidate genes associated with heat tolerance in the BAAP.

Genotypic Variation in Heat Tolerance in Aus Rice During the Flowering Stage 

Hydrogen peroxide (H2O2) is widely used as a signalling molecule. It is critical to measure H2O2 with a high spatial and temporal resolution to better understand its role as a messenger. The most frequent approach to detect H2O2 has been using chemical probes, which have the disadvantage of limited specificity and cannot report any dynamics or spatial information. Genetically encoded probes have made significant advances to overcome these limitations. Probes of the HyPer family consist of a circularly permuted yellow fluorescent protein (cpYFP) core and a sensing domain from a bacterial H2O2-activated transcription factor (OxyR). These, however, have a pronounced pH sensitivity. Other H2O2 sensors, such as roGFP2-Orp1, have shown to be pH-insensitive but lack the responsiveness to detect low nanomolar H2O2 changes, possibly due to the strong reducing effect of the glutathione redox potential (EGSH). A new ultrasensitive HyPer7 fuses the OxyR of N. meningitidis to an optimized cpYFP largely pH-independent. Here, we have generated and characterized stable Arabidopsis thaliana plants expressing a cytosolic version of HyPer7 and prove that the reduction of this sensor is mostly EGSH-independent and highly responsive to externally imposed oxidation via H2O2 feeding. Furthermore, HyPer7 could sense stress-induced H2O2 in live-cell plants exposed to herbicides, such as methyl viologen or the flagellin-induced apoplastic ROS burst, much better than the previously mentioned H2O2 probes. HyPer7 sensitivity also allowed us to see that the laser light used for excitation is sufficient to induce H2O2 release from chloroplasts in amounts only detected by this ultrasensitive sensor.

The latest HyPe(r) in plant H2O2 biosensing

The evolution of lignification in vascular plants conferred improved structural support, defence, and water transport, improving the success of plants on land. Despite lignification being a land plant specific expansion, lignin-like compounds have been identified in the cell walls of intertidal seaweeds. Here, I aim to characterise the function and structure of the ‘phycolignin’ fraction of the chlorophyte seaweed Ulva compressa. UV excitation of Ulva cell wall autofluorescence using confocal microscopy induced a blue-green emission profile comparable to that of lignin, which was significantly elevated under hyposaline stress (1.7X fold change to control, ***) and significantly reduced under hypersaline stress (0.5X fold change, ***), indicating a possible osmoprotective role. Cell walls were then isolated and fractionated into pectin, hemicellulose, and ⍺-cellulose to identify this fluorophore, with plant biomass also fractionated as a ‘lignin positive’ control. Several biochemical tests were performed to identify which polysaccharides lignin-like structures were associated with, followed by ATR-FTIR and NMR. ‘Phycolignin’ in Ulva appeared to be associated with cellulose and hemicellulose, with potential aromatic groups identified by ATR-FTIR and 1H NMR (solution-state). Positive identification of lignin within plant extracts showed that Ulva ‘phycolignin’ displayed fundamental structural differences to true lignin. However, the lack of available reference spectra for FTIR of chlorophyte phenolics, and the general insolubility of the Ulva cell wall in conventional NMR solvents limited data interpretation. Better structural characterisation of ‘phycolignin’ in Ulva will increase our understanding on how multicellularity, the metabolome, and stress responses evolved and diverge within the green lineage.

Characterisation of 'phycolignin' in the cell wall of chlorophyte seaweeds

We demonstrate that one-pot consolidated bioprocessing using deep eutectic solvents can improve the extraction of fermentable sugars from the green seaweed, Ulva linza. Ulva biomass was first pretreated with either sulfuric acid, the deep eutectic solvent choline chloride-oxalic acid, or the ionic liquid 1-ethyl-3-methylimidazolium bromide. After pretreatment, we separated solubilised sugars from residual solids. The residual solids were saccharified using green solvent-stable glycosidases that were isolated from a newly identified strain of the ascomycetous fungus, Penicillium oxalicum. The factors affecting this saccharification were quantified from the response surfaces of a Box-Behnken design. Finally, the sugars extracted from Ulva biomass were fermented into bioethanol using a dual yeast culture. We obtained higher bioethanol yields from Ulva biomass with consolidated processing in deep eutectic solvents, compared to lower yields from Ulva biomass with other solvents or two-stage processing.

One-pot processing of Ulva seaweed biomass with deep eutectic solvent-optimized fermentation

Although there is much controversy regarding the effects of ocean acidification on the behaviour of marine fish, marine invertebrates have been much less studied. Dungeness crabs, Metacarcinus magister, are found on the West coast of North American and their commercial and recreational fishery is one of the most valuable fishery in the Pacific region, with annual revenues of $220 million. Dungeness crabs live in the deep-sea and enter intertidal areas in order to find food. Previous studies have found that they routinely encounter pCO2 concentrations of up to 1300 µatm in their coastal habitat and that these levels of ocean acidification can affect larval development and mechanoreceptors. We exposed adult Dungeness crabs to control (600 µatm) and high pCO2 (1300 µatm) for 10 days and assessed their behaviour and physiology using electrophysiology, gene expression, immunohistochemistry and transmission electron microscopy of their olfactory organs, the antennules. Dungeness crabs exposed to high pCO2 had a 25% lower rate of antennular flicking rate in response to all concentrations of cadaverine, a food cue, than crabs exposed to control pCO2. No differences in antennular flicking were found in response to ammonium chloride, a deterant, however there was a large increase in the expression of Rh proteins that function as ammonia transporters in their antennules indicating a potential compensatory mechanism. We propose that a more holistic approach to the overall question of the effects of ocean acidification needs to be taken at multiple levels of biological organization and using environmentally relevant odorants and concentrations.

The effects of simulated ocean acidification on the behaviour and physiology of the economically important Dungeness crab, Metacarcinus magister

The basket of nuclear pore complex (NPC) is an assembly of flexible protein filaments and acts as a dynamic scaffold at the nucleoplasmic side that regulates various essential cellular processes, including nuclear transport, RNA processing, chromatin organization, and gene expression. However, the mechanisms behind the regulation are still poorly understood, especially in plants. Here, we identified a guanylate-binding protein (GBP)-like GTPase (GBPL3) as a specific component of the NPC basket in Arabidopsis. GBPL3 is located at the nuclear rim and confers robust physical and genetic interactions with various NPC and nucleoskeleton constituents. Proximity labeling proteomics experiment suggests that GBPL3 is predominantly distributed at the NPC basket and that it is intimately associated with chromatin remodelers, transcription regulators, and RNA processing machinery. Importantly, we found that GBPL3 promotes the formation of biomolecular condensates at the NPC basket, which is capable of recruiting its interactors. Loss of GBPL3 led to severely stunted plant growth, aberrant transcription of abiotic and biotic stress-related genes, and retention of total mRNA in the nucleus. These results suggest that GBPL3 is a plant-specific NPC basket component and can recruit transcription and RNA processing machinery through condensate formation to modulate gene expression and subsequent RNA processing, which is required for plant growth and stress responses.

GBPL3 is a novel plant nucleoporin that recruits RNA processing and transcription machinery at the nuclear basket by promoting condensates formation

Using examples of insect-plant interactions, I will show how biomechanical research can provide important insights into ecological and evolutionary questions. Insect-plant interactions must always be studied at least partially in the field, as essential aspects of their biology can be overlooked in the laboratory. Biomechanical factors play an important role in many insect-plant interactions such as pollination, herbivory and plant carnivory. Similar to the effects of plant chemical compounds or optical cues, plants can use mechanical ways to manipulate insect behaviour. For example, rockroses can mechanically detect visiting insects and guide pollinators to as yet unpollinated flowers, hollyhocks deploy slippery petals to guide insects to the fertile flower parts, and some pitcher plants have developed climbing barriers that prevent insects from walking around the pitcher to direct them to the slippery trapping surfaces. Again analogous to chemical factors in interactions, biomechanical factors can lead to adaptations and counter-adaptations, with far-reaching ecological and evolutionary consequences. The complexity of biological effects resulting from biomechanical factors is illustrated by Macaranga ant-plant mutualisms, where 'wax barriers' and 'wax-running' ants have triggered a whole range of secondary secondary ant and plant adaptive traits, including the presentation of food, epidermis longevity, the presence of preformed entrance holes, and ant colony density and aggressiveness. 'Mechanical ecology' – the study of organisms in their natural habitat with a biomechanical perspective - is not only important to elucidate adaptations of organisms, but also allows a broader search for new phenomena and organisms that could serve as inspiration for biomimetics.

Mechanical ecology of insect-plant interactions

Elucidating transcriptional regulatory networks requires knowledge of transcription factor (TF) binding sites and their occupancy, ideally in a native chromatin context. Towards this goal, we have adapted the use of micrococcal nuclease (MNase) as a biochemical probe of fixed chromatin structure with emphasis on small DNA fragments from sub-nucleosome-sized particles. Specifically, we developed MNase-defined cistrome-Occupancy Analysis (MOA-seq) to achieve high-resolution genome-wide detection of TF-bound sites within accessible chromatin regions, illustrated with a proof of concept study on field-grown maize ear shoots (Savadel, Hartwig, et al., 2021; PLOS Genetics). Fragment center plotting focused the footprint detection down to small particles with an average peak size of ~30bp. These MOA footprints (MFs) exhibited (1) expression-correlated associations with gene promoters, (2) enrichment at known TF-binding sites, (3) enrichment at known long-range chromatin interaction sites, and (4) enrichment at known open chromatin regions defined by other non-MNase assays such as ATAC-seq. Motif analysis of 145,000 maize earshoot MFs led to the detection of 215 sequence motif families, which were merged into non-overlapping regions numbering just above 100,000 loci and covering just below 0.1% of the maize genome. We also used MOA-seq to examine genomic response to hypoxia in human cells. From these initial studies, we conclude that MOA-seq is ideal for genome-wide identification of cistrome occupancy at high resolution using a single antibody-free assay. The technology adds a valuable method to the genomic toolkit for pinpointing cis-regulatory candidates in agriculture and medicine.

MNase-based MOA-Seq Pinpoints Native Cistrome Occupancy at 30-bp Resolution Amid Accessible Chromatin

Migratory flight requires birds to sustain high levels of energy expenditure, fueled almost exclusively by fat, for prolonged periods of time. Beyond fueling flight, the types of dietary fatty acids may influence performance. The natural doping hypothesis proposes that shorebirds feed on diets rich in essential n-3 long chain polyunsaturated fatty acids (LCPUFA), which help prime the flight muscles for endurance flight by enhancing membrane fludity, promoting rapid fatty acid transport, and activating cellular signalling pathways to increase aerobic and fatty acid oxidation capacity. We fed western sandpipers (Calidris mauri) formulated diets low or high in n-3 LCPUFA (EPA and DHA) and flew them for up to 8 hours in a wind tunnel to assess endurance flight performance. We found that a diet high in n-3 LCPUFA decreased flight energy expenditure by 12%. Activities of key oxidative enzymes, citrate synthase and carnitine palmitoyl transferase were significantly greater in the flight muscles of birds fed the n-3 LCPUFA enriched diet, and we are investigating further effects on the regulation of lipid metabolism and oxidative capacity. In addition, we present evidence for limited capacity for endogenous synthesis of n-3 LCPUFA from precursor fatty acids, suggesting an essential dietary requirement for n-3 LCPUFA. These findings support the natural doping hypothesis by demonstrating performance-enhancing effects at the biochemical and whole animal level, and they indicate a critical role of dietary n-3 LCPUFA during migration for shorebirds.

Underline digital video library

A Graph Fusion Approach for Cross-Lingual Machine Reading Comprehension

Script, Language, and Labels: Overcoming Three Discrepancies for Low-Resource Language Specialization

Magneto ionic and electrostatic generation of non

Genotypic Variation in Heat Tolerance in Aus Rice During the Flowering Stage

The latest HyPe(r) in plant H2O2 biosensing

Characterisation of 'phycolignin' in the cell wall of chlorophyte seaweeds

One-pot processing of Ulva seaweed biomass with deep eutectic solvent-optimized fermentation

The effects of simulated ocean acidification on the behaviour and physiology of the economically important Dungeness crab, Metacarcinus magister

GBPL3 is a novel plant nucleoporin that recruits RNA processing and transcription machinery at the nuclear basket by promoting condensates formation

Mechanical ecology of insect-plant interactions

MNase-based MOA-Seq Pinpoints Native Cistrome Occupancy at 30-bp Resolution Amid Accessible Chromatin

Flight testing natural doping: Influence of dietary n-3 long chain polyunsaturated fatty acids on migratory performance in western sandpipers

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES