Growing Like a Tree: Finding Trunks From Graph Skeleton Trees.

The message-passing paradigm has served as the foundation of graph neural networks (GNNs) for years, making them achieve great success in a wide range of applications. Despite its elegance, this paradigm presents several unexpected challenges for graph-level tasks, such as the long-range problem, information bottleneck, over-squashing phenomenon, and limited expressivity. In this study, we aim to overcome these major challenges and break the conventional "node- and edge-centric" mindset in graph-level tasks. To this end, we provide an in-depth theoretical analysis of the causes of the information bottleneck from the perspective of information influence. Building on the theoretical results, we offer unique insights to break this bottleneck and suggest extracting a skeleton tree from the original graph, followed by propagating information in a distinctive manner on this tree. Drawing inspiration from natural trees, we further propose to find trunks from graph skeleton trees to create powerful graph representations and develop the corresponding framework for graph-level tasks. Extensive experiments on multiple real-world datasets demonstrate the superiority of our model. Comprehensive experimental analyses further highlight its capability of capturing long-range dependencies and alleviating the over-squashing problem, thereby providing novel insights into graph-level tasks.

Generalizability of a Machine Learning Model for Improving Utilization of Parathyroid Hormone-Related Peptide Testing across Multiple Clinical Centers.

BACKGROUND: Measuring parathyroid hormone-related peptide (PTHrP) helps diagnose the humoral hypercalcemia of malignancy, but is often ordered for patients with low pretest probability, resulting in poor test utilization. Manual review of results to identify inappropriate PTHrP orders is a cumbersome process. METHODS: Using a dataset of 1330 patients from a single institute, we developed a machine learning (ML) model to predict abnormal PTHrP results. We then evaluated the performance of the model on two external datasets. Different strategies (model transporting, retraining, rebuilding, and fine-tuning) were investigated to improve model generalizability. Maximum mean discrepancy (MMD) was adopted to quantify the shift of data distributions across different datasets. RESULTS: The model achieved an area under the receiver operating characteristic curve (AUROC) of 0.936, and a specificity of 0.842 at 0.900 sensitivity in the development cohort. Directly transporting this model to two external datasets resulted in a deterioration of AUROC to 0.838 and 0.737, with the latter having a larger MMD corresponding to a greater data shift compared to the original dataset. Model rebuilding using site-specific data improved AUROC to 0.891 and 0.837 on the two sites, respectively. When external data is insufficient for retraining, a fine-tuning strategy also improved model utility. CONCLUSIONS: ML offers promise to improve PTHrP test utilization while relieving the burden of manual review. Transporting a ready-made model to external datasets may lead to performance deterioration due to data distribution shift. Model retraining or rebuilding could improve generalizability when there are enough data, and model fine-tuning may be favorable when site-specific data is limited.

Graph-Enhanced Emotion Neural Decoding.

Brain signal-based emotion recognition has recently attracted considerable attention since it has powerful potential to be applied in human-computer interaction. To realize the emotional interaction of intelligent systems with humans, researchers have made efforts to decode human emotions from brain imaging data. The majority of current efforts use emotion similarities (e.g., emotion graphs) or brain region similarities (e.g., brain networks) to learn emotion and brain representations. However, the relationships between emotions and brain regions are not explicitly incorporated into the representation learning process. As a result, the learned representations may not be informative enough to benefit specific tasks, e.g., emotion decoding. In this work, we propose a novel idea of graph-enhanced emotion neural decoding, which takes advantage of a bipartite graph structure to integrate the relationships between emotions and brain regions into the neural decoding process, thus helping learn better representations. Theoretical analyses conclude that the suggested emotion-brain bipartite graph inherits and generalizes the conventional emotion graphs and brain networks. Comprehensive experiments on visually evoked emotion datasets demonstrate the effectiveness and superiority of our approach.

Overexpression of a GIPC glycosyltransferase gene, OsGMT1, suppresses plant immunity and delays heading time in rice.

Glycosylinositol phosphorylceramides (GIPCs) are the major sphingolipids in the plant plasma membrane. In Arabidopsis, mutations of genes involved in the synthesis of GIPCs affect many physiological aspects of plants, including growth, pollen fertility, defense, and stress signaling. Loss of function of the GIPC MANNOSYL-TRANSFERASE1 (AtGMT1) results in GIPC misglycosylation and induces plant immune responses accompanied by a severely dwarfed phenotype, thus indicating that GIPCs play important roles in plant immunity. Here, we investigated the enzymatic activity and phenotypes of transgenic lines of OsGMT1, the ortholog of AtGMT1. Sphingolipidomic analysis indicated that OsGMT1 retained the enzymatic activity of GIPC hexose (Hex) glycosylation, but the knockout lines did not accumulate H2O2. In contrast, the OsGMT1 overexpression lines showed significant down-regulation of several defense-associated or cell wall synthesis-associated genes, and enhanced sensitivity to rice blast. Furthermore, we first demonstrated the sensitivity of rice cells to MoNLP1 protein through calcein AM release assays using rice protoplasts, thus legitimizing the presence of MoNLPs in rice blast fungus. In addition, yeast two-hybrid screens using OsGMT1 as bait revealed that OsGMT1 may regulate heading time through the OsHAP5C signaling pathway. Together, our findings suggested clear physiological functional differentiation of GMT1 orthologs between rice and Arabidopsis.

Transcriptome Analyses Indicate Significant Association of Increased Non-Additive and Allele-Specific Gene Expression with Hybrid Weakness in Rice (Oryza sativa L.).

The heterosis in hybrid rice is highly affected by the environment and hybrid weakness occurs frequently depending on the genotypes of the hybrid and its parents. Hybrid weakness was also observed in our field experiments on nine rice hybrids produced by 3 × 3 incomplete diallel crosses. Among the nine hybrids, five displayed mid-parent heterosis (MPH) for grain yield per plant, while four showed mid-parent hybrid weakness (MPHW). A sequencing analysis of transcriptomes in panicles at the seed-filling stage revealed a significant association between enhanced non-additive gene expression (NAE) and allele-specific gene expression (ASE) with hybrid weakness. High proportions of ASE genes, with most being of mono-allele expression, were detected in the four MPHW hybrids, ranging from 22.65% to 45.97%; whereas only 4.80% to 5.69% of ASE genes were found in the five MPH hybrids. Moreover, an independence test indicated that the enhancements of NAE and ASE in the MPHW hybrids were significantly correlated. Based on the results of our study, we speculated that an unfavorable environment might cause hybrid weakness by enhancing ASE and NAE at the transcriptome level.

RNA-seq profiling of primary calli induced by different media and photoperiods for japonica rice 'Yunyin'.

The induction of embryogenic calli plays a vital role in the genetic transformation and regeneration of rice (Oryza sativa L.). Despite progress in rice tissue culture, the molecular mechanisms of embryogenic callus induction remain unknown. In this study, gene expression profiles associated with calli were comprehensively analyzed during callus induction of japonica rice 'Yunyin'. We first confirmed that NMB medium with 24 h of light and 0 h of dark (NMB-L) was the optimal condition for 'Yunyin' callus induction, while J3 medium with 0 h of light and 24 h of dark (J3-D) was the worst condition. After transcriptome analysis, 33,597 unigenes were assembled, among which we identified 6,063 DEGs (Differentially Expressed Genes) related to media and seven DEGs related to photoperiod. Phenylpropanoid biosynthesis, plant hormone signal, and starch and sucrose metabolism were the top three pathways affected by media, while the circadian rhythm-plant pathway was associated with photoperiod. Furthermore, we identified two candidate genes, Os01g0965900 and Os12g0555200, affected by both medium and photoperiod. Statistical analysis of RNA-seq libraries showed that the expression levels of these two genes in J3-D calli were over 2.5 times higher than those in NMB-L calli, which was further proved by RT-qPCR analysis. Based on FPKM (Fragments Per Kilobase of transcript Per Million mapped reads), unigenes belonging to the NMB-L group were mainly assigned to ribosome, carbon metabolism, biosynthesis of amino acids, protein processing in endoplasmic reticulum, and plant hormone signal transduction pathways. We transformed Os12g0555200Nip and Os12g05552009311 into 'Nipponbare' calli and observed their effects on the growth and development process of rice calli using TEM (Transmission Electron Microscopy) and SEM (Scanning Electron Microscopy). Observations showed that Os12g05552009311 was more disadvantageous to rice callus growth than Os12g0555200Nip. Our results reveal that the Os12g0555200, identified from transcriptomic profiles, has a negative influence during 'Yunyin' callus induction. Supplementary information: The online version contains supplementary material available at 10.1007/s11032-022-01283-y.