Three critical factors affecting automated image species recognition performance for camera traps

Ecological camera traps are increasingly used by wildlife biologists to unobtrusively monitor an ecosystems animal population. However, manual inspection of the images produced is expensive, laborious, and time‐consuming. The success of deep learning systems using camera trap images has been previously explored in preliminary stages. These studies, however, are lacking in their practicality. They are primarily focused on extremely large datasets, often millions of images, and there is little to no focus on performance when tasked with species identification in new locations not seen during training. Our goal was to test the capabilities of deep learning systems trained on camera trap images using modestly sized training data, compare performance when considering unseen background locations, and quantify the gradient of lower bound performance to provide a guideline of data requirements in correspondence to performance expectations. We use a dataset provided by Parks Canada containing 47,279 images collected from 36 unique geographic locations across multiple environments. Images represent 55 animal species and human activity with high‐class imbalance. We trained, tested, and compared the capabilities of six deep learning computer vision networks using transfer learning and image augmentation: DenseNet201, Inception‐ResNet‐V3, InceptionV3, NASNetMobile, MobileNetV2, and Xception. We compare overall performance on “trained” locations where DenseNet201 performed best with 95.6% top‐1 accuracy showing promise for deep learning methods for smaller scale research efforts. Using trained locations, classifications with <500 images had low and highly variable recall of 0.750 ± 0.329, while classifications with over 1,000 images had a high and stable recall of 0.971 ± 0.0137. Models tasked with classifying species from untrained locations were less accurate, with DenseNet201 performing best with 68.7% top‐1 accuracy. Finally, we provide an open repository where ecologists can insert their image data to train and test custom species detection models for their desired ecological domain.     

Biomolecular condensates in membrane receptor signaling

Clustering is a prominent feature of receptors at the plasma membrane (PM). It plays an important role in signaling. Liquid–liquid phase separation (LLPS) of proteins is emerging as a novel mechanism underlying the observed clustering. Receptors/transmembrane signaling proteins can be core components essential for LLPS (such as LAT or nephrin) or clients enriched at the phase-separated condensates (for example, at the postsynaptic density or at tight junctions). Condensate formation has been shown to regulate signaling in multiple ways, including by increasing protein binding avidity and by modulating the local biochemical environment. In moving forward, it is important to study protein LLPS at the PM of living cells, its interplay with other factors underlying receptor clustering, and its signaling and functional consequences.     

Plasticity of the skeleton and skeletal deformities in zebrafish (Danio rerio) linked to rearing density

The teleost zebrafish (Danio rerio), an established model for human skeletal diseases, is reared under controlled conditions with defined parameters for temperature and photoperiod. Studies aimed at defining the proper rearing density have been performed with regard to behavioural and physiological stress response, sex ratio and reproduction. Studies concerning the effect of rearing density on the skeletal phenotype are lacking. This study analyses the response of the skeleton to different rearing densities and describes the skeletal deformities. Wild-type zebrafish were reared up to 30 dpf (days post-fertilization) in a common environment. From 30 to 90 dpf, animals were reared at three different densities: high density (HD), 32 fish l^–1; medium density (MD), 8 fish l^–1 and low density (LD), 2 fish l^–1. Animals at 30 and 90 dpf were collected and whole-mount stained with Alizarin red S to visualize mineralized tissues. The entire skeleton was analysed for meristic counts and 172 types of deformities. The results showed that the rearing density significantly influenced the specimens’ average standard length, which decreased with the increase in the rearing density. Differences in meristic counts among the three groups were not observed. Rearing density–independent malformations affected the ribs, neural arches and the spines of the abdominal region, as well as vertebrae of the caudal complex. The HD group showed the highest number of deformities per specimen, the highest number of observed types of deformities and, together with the MD group, the highest frequency of specimens affected by severe deformities. In particular, the HD group showed deformities affecting arches, spines and vertebral centra in the caudal region of the vertebral column. This study provides evidence of an effect of the rearing density on the development of different skeletal phenotypes.     

Maternal genetic history of southern East Asians over the past 12,000 years

Southern East Asia, including Guangxi and Fujian provinces in China, is home to diverse ethnic groups, languages, and cultures. Previous studies suggest a high complexity regarding population dynamics and the history of southern East Asians. However, large-scale genetic studies on ancient populations in this region are hindered by limited sample preservation. Here, using highly efficient DNA capture techniques, we obtain 48 complete mitochondrial genomes of individuals from Guangxi and Fujian in China and reconstruct their maternal genetic history over the past 12,000 years. We find a strong connection between southern East Asians dating to ~12,000–6000 years ago and present-day Southeast Asians. In addition, stronger genetic affinities to northern East Asians are observed in historical southern East Asians than Neolithic southern East Asians, suggesting increased interactions between northern and southern East Asians over time. Overall, we reveal dynamic connections between ancient southern East Asians and populations located in surrounding regions, as well as a shift in maternal genetic structure within the populations over time.     

Genome assisted molecular typing and pathotyping of rice blast pathogen,Magnaporthe oryzae,reveals a genetically homogenous population with high virulence diversity

Genome sequence-driven molecular typing tools have the potential to uncover the population biology and genetic diversity of rapidly evolving plant pathogens like Magnaporthe oryzae. Here, we report a new molecular typing technique -a digitally portable tool for population genetic analysis of M. oryzae to decipher the genetic diversity. Our genotyping tool exploiting allelic variations in housekeeping and virulence genes coupled with pathotyping revealed a prevalence of genetically homogenous populations within a single-field and plant niches such as leaf and panicle. The M. oryzae inciting leaf-blast and panicle-blast were confirmed to be genetically identical with no or minor nucleotide polymorphism in 17 genomic loci analyzed. Genetic loci such as Mlc1, Mpg1, Mps1, Slp1, Cal, Ef-Tu, Pfk, and Pgk were highly polymorphic as indicated by the haplotype-diversity, the number of polymorphic sites, and the number of mutations. The genetically homogenous single field population showed high virulence variability or diversity on monogenic rice differentials. The study indicated that the genetic similarity displayed by the isolates collected from a particular geographical location had no consequence on their virulence pattern on rice differentials carrying single/multiple resistance genes. The data on virulence diversity showed by the identical Sequence Types (STs) is indicative of no congruence between polymorphic virulence genes-based pathotyping and conserved housekeeping genes-based genotyping.     

Uncovering the different scales in deer–forest interactions

Deer are regarded to be a keystone species as they play a crucial role in the way an ecosystem functions. Most deer–forest interaction studies apply a single scale — process of analyzing ecological interactions by only taking into account one dependent variable — to understand how deer browsing behavior shapes different forest components, but they overlook the fact that forests respond to multiple scales simultaneously. This research evaluates the effect of browsing by wild deer on temperate and boreal forests at different scales by synthesizing seminal papers, specifically (a) what are the effects of deer population density in forest regeneration? (b) What are the effects of deer when forests present diverging spatial characteristics? (c) What are the effects on vegetation at different temporal scales? and (d) What are the hierarchical effects of deer when considering other trophic levels? Additionally, a framework based on modern technology is proposed to answer the multiscale research questions previously identified. When analyzing deer–forest interactions at different scales, the strongest relationships occur at the extremes. For example: when deer assemblage occurs in low or high density and is composed of a mix of small and large species. As forests on poor soils remain restrained in size, isolated and chronically browsed. When forests harbor incomplete trophic levels, the effects spill over to lower trophic levels. To better understand the complexities in deer–forest interactions, researchers should combine technology‐based instruments like fixed sensors and drones with field‐tested methods such observational studies and experiments to tackle multiscale research questions.     

Age-related increase of β1-adrenergic receptor gene expression in rat liver: a potential mechanism contributing to increased β-adrenergic receptor density and responsiveness during aging

In this study we examined whether the levels of gene expressions of the three β- adrenergic receptor (βAR) subtypes, β₁, β₂, and β₃, contribute to age-related increase in βAR density. Liver membranes and total RNA were prepared from young (4- to 6-month-old) and old (24-month-old) male Fischer 344 rats. βAR density (B_max) in liver membranes was measured by a radioligand receptor binding assay using the receptor subtype nonselective βAR antagonist ¹²⁵I-pindolol as the radioligand. Steady-state levels of β₂AR mRNA in rat liver were measured by Northern blot analysis; because of the low abundance of β₁AR and β₃AR mRNA in rat liver, the expressions of these genes were measured by a semiquantitative RT-PCR or an RT-PCR. Scatchard analysis of saturation binding curves of the binding assay confirmed an age-related increase in B_max (young: 7.1?±?0.8?fmol/mg protein vs. old: 18.1?±?4.3?fmol/mg protein). No age-related differences were found in the levels of β₂AR mRNA. However, semiquantitative RT-PCR revealed an approximately twofold increase in β₁AR mRNA level between young and old rats (P?<?0.05). β₁AR mRNA levels were also correlated with B_max values for ¹²⁵I-pindolol binding sites in individual rats (r = 0.67; P?=?0.012). β₃AR mRNA, which was demonstrable in rat white adipose tissue by RT-PCR, was generally not detected in livers from young or old rats, with the exception of two old rats with the highest B_max. These results suggest that an age-related increase of β₁AR gene expression contributes to increased βAR density and β adrenergic responsiveness in rat liver during aging.     

Experimental and modelling studies of drought‐adaptive root architectural traits in wheat (Triticum aestivum L.)

Abstract

This paper presents an interdisciplinary approach to crop improvement that links physiology with plant breeding and simulation modelling to enhance the selection of high‐yielding, drought‐tolerant varieties. In a series of field experiments in Queensland, Australia, we found that the yield of CIMMYT wheat line SeriM82 ranged from 6% to 28% greater than the current cultivar Hartog. Physiological studies on the adaptive traits revealed that SeriM82 had a narrower root architecture and extracted more soil moisture, particularly deep in the profile. Results of a simulation analysis of these adaptive root traits with the cropping system model APSIM for a range of rain‐fed environments in southern Queensland indicated a mean relative yield benefit of 14.5% in water‐deficit seasons. Furthermore, each additional millimetre of water extracted during grain filling generated an extra 55 kg ha^?1 of grain yield. Further root studies of a large number of wheat genotypes revealed that wheat root architecture is closely linked to the angle of seminal roots at the seedling stage – a trait which is suitable for large‐scale and cost‐effective screening programmes. Overall, our results suggest that an interdisciplinary approach to crop improvement is likely to enhance the rate of yield improvement in rain‐fed crops.