RESUMO
Understanding the function and fitness effects of diverse plant genomes requires transferable models. Language models (LMs) pre-trained on large-scale biological sequences can learn evolutionary conservation, thus expected to offer better cross-species prediction through fine-tuning on limited labeled data compared to supervised deep learning models. We introduce PlantCaduceus, a plant DNA LM based on the Caduceus and Mamba architectures, pre-trained on a carefully curated dataset consisting of 16 diverse Angiosperm genomes. Fine-tuning PlantCaduceus on limited labeled Arabidopsis data for four tasks involving transcription and translation modeling demonstrated high transferability to maize that diverged 160 million years ago, outperforming the best baseline model by 1.45-fold to 7.23-fold. PlantCaduceus also enables genome-wide deleterious mutation identification without multiple sequence alignment (MSA). PlantCaduceus demonstrated a threefold enrichment of rare alleles in prioritized deleterious mutations compared to MSA-based methods and matched state-of-the-art protein LMs. PlantCaduceus is a versatile pre-trained DNA LM expected to accelerate plant genomics and crop breeding applications.
RESUMO
Prostate cancer (PCa) is the most common malignancy of the male urinary system. Mitophagy, as a type of autophagy, can remove damaged mitochondria in cells. Mitophagy-related genes (MRGs) have been shown to play critical roles in the development of PCa. To this end, based on the comprehensive analysis of RNA-seq and scRNA-seq data of PCa samples and their controls, this paper identified PCa subtypes and constructed a prognostic model. In this paper, we downloaded scRNA-seq and RNA-seq data from Gene Expression Omnibus (GEO) and TCGA database. Based on the R package "Seurat" to process the scRNA-seq data, a total of five cell types were identified. Each cell population was scored based on the R package "AUCell" and using the intersection genes between MRGs and each cell population. The B cell population was then identified as a high-scoring cell population. Differentially expressed genes in RNA-seq data were identified based on the R package "limma" and intersected with previously intersected genes. Then, based on univariate Cox regression analysis and Lasso-Cox regression analysis, the prognostic genes were screened, and the risk model was constructed (composed of ADH5, CAT, BCAT2, DCXR, OGT, and FUS). The model is validated on internal and external test sets. Independent prognostic analysis identified age, N stage, and risk score as independent prognostic factors. This paper's risk models and prognostic genes can provide a reference for developing novel therapeutic targets for PCa.
RESUMO
Long-non-coding RNAs (lncRNAs) are defined as RNA sequences which are >200 nt with no coding capacity. These lncRNAs participate in various biological mechanisms, and are widely abundant in a diversity of species. There is well-documented evidence that lncRNAs can interact with genomic DNAs by forming triple helices (triplexes). Previously, several computational methods have been designed based on the Hoogsteen base-pair rule to find theoretical RNA-DNA:DNA triplexes. While powerful, these methods suffer from a high false-positive rate between the predicted triplexes and the biological experiments. To address this issue, we first collected the experimental data of genomic RNA-DNA triplexes from antisense oligonucleotide (ASO)-mediated capture assays and used Triplexator, the most widely used tool for lncRNA-DNA interaction, to reveal the intrinsic information on true triplex binding potential. Based on the analysis, we proposed six computational attributes as filters to improve the in-silico triplex prediction by removing most false positives. Further, we have built a new database, TRIPBASE, as the first comprehensive collection of genome-wide triplex predictions of human lncRNAs. In TRIPBASE, the user interface allows scientists to apply customized filtering criteria to access the potential triplexes of human lncRNAs in the cis-regulatory regions of the human genome. TRIPBASE can be accessed at https://tripbase.iis.sinica.edu.tw/.
RESUMO
The cerebral ischemia-reperfusion model was established to investigate the anti-oxidation and anti-apoptosis mechanism of Xinshao formula on the cerebral ischemia reperfusion injury in rats. SD rats were randomly divided into five groups: sham operation group, model group, and low, middle and high-dose Xinshao formula groups (0.31, 0.62, 1.25 g·kg⻹). After administration with Xinshao formula for 7 days, the rats were used to establish the cerebral ischemia-reperfusion model. The neurological behavior was evaluated. TTC staining was implemented to determine the volume of cerebral infarction. The levels of ROS, SOD, GSH-PX, NO and iNOS in serum were examined, and the mRNA and protein levels of Bcl-2, Bax and caspase 3 in hippocampal CA1 were detected by qRT-PCR, immunohistochemical assay and Western blot. It is found that Xinshao formula could significantly reduce the symptoms of nervous function and the volume of cerebral infarction in MACO rats. Compared with model group, the rats in Xinshao formula group showed increases in the activities of SOD and GSH-PX (P<0.01), and decreases in the activities of iNOS and the contents of NO, ROS and MDA (P<0.01). In addition, Xinshao formula could down-regulate the mRNA and protein levels of Bax and caspase 3 (P<0.01), and up-regulated those of Bcl-2 (P<0.01) in MACO rats. In conclusion, Xinshao formula showed aprotective effect on cerebral ischemia reperfusion injuryin rats, which may be associated with the promotion of anti-oxidation and anti-apoptosis.
