Composite Object Relation Modeling for Few-Shot Scene Recognition.

The goal of few-shot image recognition is to classify different categories with only one or a few training samples. Previous works of few-shot learning mainly focus on simple images, such as object or character images. Those works usually use a convolutional neural network (CNN) to learn the global image representations from training tasks, which are then adapted to novel tasks. However, there are many more abstract and complex images in real world, such as scene images, consisting of many object entities with flexible spatial relations among them. In such cases, global features can hardly obtain satisfactory generalization ability due to the large diversity of object relations in the scenes, which may hinder the adaptability to novel scenes. This paper proposes a composite object relation modeling method for few-shot scene recognition, capturing the spatial structural characteristic of scene images to enhance adaptability on novel scenes, considering that objects commonly co- occurred in different scenes. In different few-shot scene recognition tasks, the objects in the same images usually play different roles. Thus we propose a task-aware region selection module (TRSM) to further select the detected regions in different few-shot tasks. In addition to detecting object regions, we mainly focus on exploiting the relations between objects, which are more consistent to the scenes and can be used to cleave apart different scenes. Objects and relations are used to construct a graph in each image, which is then modeled with graph convolutional neural network. The graph modeling is jointly optimized with few-shot recognition, where the loss of few-shot learning is also capable of adjusting graph based representations. Typically, the proposed graph based representations can be plugged in different types of few-shot architectures, such as metric-based and meta-learning methods. Experimental results of few-shot scene recognition show the effectiveness of the proposed method.

Polymorphisms in 5' proximal regulating region of THRSP gene are associated with fat production in pigs.

Chinese and imported pig breeds differ in fat production potential, which is associated with the polymorphisms in the 5' proximal regulating region (5'PRR) of thyroid hormone responsive gene (THRSP). In three Chinese breeds (Dingyuan, CDY; Wannanhua, CWH; and Jixi, CJX) and one introduced breed (Yorkshire, YKS), three variant sites were located at T/C-400, A/G-376, and G/A-98 in the 5'PRR. Chinese pig breeds had higher C-400 allele frequencies than YKS. The frequencies of A-376 in CDY and G-376 in CWH were about 0.8. G-98 allele frequencies in CWH and YKS were 0.8617 and 0.8149, respectively. TGG was the dominant haplotype in YKS, CGG in CWH and CJX, and CAA in CDY. According to haplotype frequency, four breeds were clustered into three types, which was consistent with the geographical distribution of the breeds. In CDY, the average backfat thickness (BFT) was the highest with the CC-400 genotype, followed by CT-400 and TT-400 genotypes. In YKS, the pigs with CC-400 or CT-400 genotypes had higher BFT and average daily weight gain, whereas those with CC-400 or TT-400 genotypes had larger lion-eye area. No significant difference was observed in carcass traits among different genotypes at the A/G-376 and G/A-98 loci. The mRNA abundance of THRSP expression for the CCAGAG genotype was significantly higher than that for CTAGAG or TTAGAG genotype. These results indicated that the polymorphisms and genotype distribution of THRSP were closely related to the potential for fat production in pig breeds, which were the result of adaptation to artificial selection and natural selection.

Genome-Wide Identification, Classification, and Expression Profiling Reveals R2R3-MYB Transcription Factors Related to Monoterpenoid Biosynthesis in Osmanthus fragrans.

Osmanthus fragrans is widely grown for the purpose of urban greening and the pleasant aroma emitted from its flowers. The floral scent is determined by several monoterpenoid volatiles, such as linalool and its oxides, which are a few of the most common volatiles and the main components of the essential oils in most sweet osmanthus cultivars. In addition, the relative contents of cis- and trans-linalool oxide (furan) may affect the aromas and quality of the essential oils. MYB proteins represent the largest family of transcription factors in plants and participate in regulating secondary metabolites. Several cis-elements, especially AC-rich regions, are known to be bound by 2R-MYBs and could be found in the promoter of the enzyme genes in the terpenoid metabolic pathway. However, there has to date been no investigation into the 2R-MYB family genes involved in regulating terpenoid biosynthesis in O. fragrans. Here, 243 non-redundant 2R-MYB proteins were grouped into 33 clusters based on the phylogeny and exon-intron distribution. These genes were unevenly distributed on 23 chromosomes. Ka/Ks analysis showed that the major mode of 2R-MYB gene evolution was purifying selection. Expression analysis indicated that 2R-MYB genes in O. fragrans exhibited varied expression patterns. A total of 35 OfMYBs representing the highest per kilobase per million mapped reads in the flower were selected for quantitative real-time PCR analysis. The correlation analysis between the expression level and the contents of fragrant compounds at different flowering stages suggested that OfMYB19/20 exhibited remarkably positive correlation with the accumulation of cis-linalool oxides. OfMYB51/65/88/121/137/144 showed significantly negative correlations with one or more linalool oxides. Characterization of these proteins revealed that OfMYB19 and OfMYB137 were localized in the nuclei, but did not show transcriptional activation in the yeast system, which suggested that they may be bound to other transcription factors to exert regulatory functions. These findings provide useful information for further functional investigation of the 2R-MYBs and offer a foundation for clarifying the 2R-MYB transcription factors involved in the molecular mechanism of the regulation of monoterpenoid biosynthesis in Osmanthus fragrans.

Scene Recognition with Prototype-agnostic Scene Layout.

Exploiting the spatial structure in scene images is a key research direction for scene recognition. Due to the large intra-class structural diversity, building and modeling flexible structural layout to adapt various image characteristics is a challenge. Existing structural modeling methods in scene recognition either focus on predefined grids or rely on learned prototypes, which all have limited representative ability. In this paper, we propose Prototype-agnostic Scene Layout (PaSL) construction method to build the spatial structure for each image without conforming to any prototype. Our PaSL can flexibly capture the diverse spatial characteristic of scene images and have considerable generalization capability. Given a PaSL, we build Layout Graph Network (LGN) where regions in PaSL are defined as nodes and two kinds of independent relations between regions are encoded as edges. The LGN aims to incorporate two topological structures (formed in spatial and semantic similarity dimensions) into image representations through graph convolution. Extensive experiments show that our approach achieves state-of-the-art results on widely recognized MIT67 and SUN397 datasets without multi-model or multi-scale fusion. Moreover, we also conduct the experiments on one of the largest scale datasets, Places365. The results demonstrate the proposed method can be well generalized and obtains competitive performance.

SSR marker development in Clerodendrum trichotomum using transcriptome sequencing.

Clerodendrum trichotomum, a member of the Lamiaceae (Verbenaceae) family, is an ornamental plant widely distributed in South Asia. Previous studies have focused primarily on its growth characteristics, stress resistance, and pharmacological applications; however, molecular investigations remain limited. Considering germplasm conservation and the extensive applications of this plant, it is necessary to explore transcriptome resources and SSR makers for C. trichotomum. In the present study, RNA sequencing was used to determine the transcriptome of C. trichotomum. Subsequently, unigene annotations and classifications were obtained, and SSRs were mined with MIcroSAtellite. Finally, primer pairs designed with Oligo 6.0 were selected for polymorphism validation. In total, 127,325,666 high-quality reads were obtained, and 58,345 non-redundant unigenes were generated, of which 36,900 (63.24%) were annotated. Among the annotated unigenes, 35,980 (97.51%) had significant similarity to 607 species in Nr databases. In addition, a total of 6,444 SSRs were identified in 5,530 unigenes, and 200 random primer pairs were designed for polymorphism validation. Furthermore, after primary polymorphism identification, 30 polymorphic primer pairs were selected for the further polymorphism screening, and 200 alleles were identified, 197 of which showed polymorphism. In this work, a large number of unigenes were generated, and numerous SSRs were detected. These findings should be beneficial for further investigations into germplasm conservation and various applications of C. trichotomum. These results should also provide a solid foundation for future molecular biology studies in C. trichotomum.

Erratum: Author Correction: The chromosome-level quality genome provides insights into the evolution of the biosynthesis genes for aroma compounds of Osmanthus fragrans.

[This corrects the article DOI: 10.1038/s41438-018-0108-0.].

The chromosome-level quality genome provides insights into the evolution of the biosynthesis genes for aroma compounds of Osmanthus fragrans.

Sweet osmanthus (Osmanthus fragrans) is a very popular ornamental tree species throughout Southeast Asia and USA particularly for its extremely fragrant aroma. We constructed a chromosome-level reference genome of O. fragrans to assist in studies of the evolution, genetic diversity, and molecular mechanism of aroma development. A total of over 118 Gb of polished reads was produced from HiSeq (45.1 Gb) and PacBio Sequel (73.35 Gb), giving 100× depth coverage for long reads. The combination of Illumina-short reads, PacBio-long reads, and Hi-C data produced the final chromosome quality genome of O. fragrans with a genome size of 727 Mb and a heterozygosity of 1.45 %. The genome was annotated using de novo and homology comparison and further refined with transcriptome data. The genome of O. fragrans was predicted to have 45,542 genes, of which 95.68 % were functionally annotated. Genome annotation found 49.35 % as the repetitive sequences, with long terminal repeats (LTR) being the richest (28.94 %). Genome evolution analysis indicated the evidence of whole-genome duplication 15 million years ago, which contributed to the current content of 45,242 genes. Metabolic analysis revealed that linalool, a monoterpene is the main aroma compound. Based on the genome and transcriptome, we further demonstrated the direct connection between terpene synthases (TPSs) and the rich aromatic molecules in O. fragrans. We identified three new flower-specific TPS genes, of which the expression coincided with the production of linalool. Our results suggest that the high number of TPS genes and the flower tissue- and stage-specific TPS genes expressions might drive the strong unique aroma production of O. fragrans.

Transcriptomic Analysis of the Candidate Genes Related to Aroma Formation in Osmanthus fragrans.

Osmanthus fragrans, or "RiXiangGui", is an ornamental, woody, evergreen plant that is cultivated widely because it blooms recurrently and emits a strong fragrance. Recently, the germplasm resources, classification, and aroma compositions of O. fragrans have been investigated. However, the molecular mechanisms of the floral scent formation and regulation have remained largely unknown. To obtain a global perspective on the molecular mechanism of the aroma formation during blooming, nine RNA Sequencing (RNA-Seq) libraries were constructed from three flowering stages: The initial, full, and final flowering stage. In short, a total of 523,961,310 high-quality clean reads were assembled into 136,611unigenes, with an average sequence length of 792 bp. About 47.43% of the unigenes (64,795) could be annotated in the NCBI non-redundant protein database. A number of candidate genes were identified in the terpenoid metabolic pathways and 1327 transcription factors (TFs), which showed differential expression patterns among the floral scent formation stages, were also identified, especially OfMYB1, OfMYB6, OfWRKY1, and OfWRKY3, which could play critical roles in the floral scent formation. These results indicated that the floral scent formation of O. fragrans was a very complex process which involved a large number of TFs. This study provides reliable resources for further studies of the O.fragrans floral scent formation.