OUP accepted manuscript

Inference of differential gene regulatory networks using boosted differential trees

Summary

Diseases can be caused by molecular perturbations that induce specific changes in regulatory interactions and their coordinated expression, also referred to as network rewiring. However, the detection of complex changes in regulatory connections remains a challenging task and would benefit from the development of novel nonparametric approaches. We develop a new ensemble method called BoostDiff (boosted differential regression trees) to infer a differential network discriminating between two conditions. BoostDiff builds an adaptively boosted (AdaBoost) ensemble of differential trees with respect to a target condition. To build the differential trees, we propose differential variance improvement as a novel splitting criterion. Variable importance measures derived from the resulting models are used to reflect changes in gene expression predictability and to build the output differential networks. BoostDiff outperforms existing differential network methods on simulated data evaluated in four different complexity settings. We then demonstrate the power of our approach when applied to real transcriptomics data in COVID-19, Crohn's disease, breast cancer, prostate adenocarcinoma, and stress response in Bacillus subtilis. BoostDiff identifies context-specific networks that are enriched with genes of known disease-relevant pathways and complements standard differential expression analyses.

Availability and implementation

BoostDiff is available at https://github.com/scibiome/boostdiff_inference.

Long non-coding RNA NRAV in the 12q24.31 risk locus drives gastric cancer development through glucose metabolism reprogramming

Long non-coding RNAs (lncRNAs) serve as vital candidates to mediate cancer risk. Here, we aimed to identify the risk single-nucleotide polymorphisms (SNPs)-induced lncRNAs and to investigate their roles in gastric cancer (GC) development. Through integrating the differential expression analysis of lncRNAs in GC tissues and expression quantitative trait loci analysis in normal stomach tissues and GC tissues, as well as genetic association analysis based on GC genome-wide association studies and an independent validation study, we identified four lncRNA-related SNPs consistently associated with GC risk, including SNHG7 [odds ratio (OR) = 1.16, 95% confidence interval (CI): 1.09-1.23], NRAV (OR = 1.11, 95% CI: 1.05-1.17), LINC01082 (OR = 1.16, 95% CI: 1.08-1.22) and FENDRR (OR = 1.16, 95% CI: 1.07-1.25). We further found that a functional SNP rs6489786 at 12q24.31 increases binding of MEOX1 or MEOX2 at a distal enhancer and results in up-regulation of NRAV. The functional assays revealed that NRAV accelerates GC cell proliferation while inhibits GC cell apoptosis. Mechanistically, NRAV decreases the expression of key subunit genes through the electron transport chain, thereby driving the glucose metabolism reprogramming from aerobic respiration to glycolysis. These findings suggest that regulating lncRNA expression is a crucial mechanism for risk-associated variants in promoting GC development.

Identification and immunoassay of prognostic genes associated with the complement system in acute myeloid leukemia

Studies have associated the development of pulmonary leukemia with the activation of the complement system. However, the roles and mechanisms of complement system-related genes (CSRGs) in acute myeloid leukemia (AML) have not been investigated extensively. This study used The Cancer Genome Atlas (TCGA)-AML and GSE37642 datasets. Differentially expressed CSRGs (CSRDEGs) were identified by overlapping genes differentially expressed between the high and low CSRG score groups and key module genes identified in a weighted gene co-expression network analysis. Univariate and multivariate Cox analyses identified CSRG-related biomarkers, which were used to build a prognostic model. After gene set enrichment analysis (GSEA), immune-related and drug-sensitivity analyses were performed in the high- and low-risk groups. Four prognosis-related biomarkers were identified and used to develop a prognostic model: MEOX2, IGFBP5, CH25H, and RAB3B. The model's performance was verified in a test cohort (a subset of samples from the TCGA-AML dataset) and a validation cohort (GSE37642). The GSEA revealed that the high-risk group was mainly enriched for Golgi organization and cytokine-cytokine receptor interactions, and the low-risk group was mainly enriched in the hedgehog signaling pathway and spliceosome. Lastly, two immune cells were found to show differential infiltration between risk groups, which correlated with the risk scores. M1 macrophage infiltration was significantly positively correlated with RAB3B expression. Sensitivity to 36 drugs differed significantly between risk groups. This study screened four CSRG-related biomarkers (MEOX2, IGFBP5, CH25H, and RAB3B) to provide a basis for predicting AML prognosis.

Applications of organoid technology to brain tumors

Lacking appropriate model impedes basic and preclinical researches of brain tumors. Organoids technology applying on brain tumors enables great recapitulation of the original tumors. Here, we compared brain tumor organoids (BTOs) with common models including cell lines, tumor spheroids, and patient-derived xenografts. Different BTOs can be customized to research objectives and particular brain tumor features. We systematically introduce the establishments and strengths of four different BTOs. BTOs derived from patient somatic cells are suitable for mimicking brain tumors caused by germline mutations and abnormal neurodevelopment, such as the tuberous sclerosis complex. BTOs derived from human pluripotent stem cells with genetic manipulations endow for identifying and understanding the roles of oncogenes and processes of oncogenesis. Brain tumoroids are the most clinically applicable BTOs, which could be generated within clinically relevant timescale and applied for drug screening, immunotherapy testing, biobanking, and investigating brain tumor mechanisms, such as cancer stem cells and therapy resistance. Brain organoids co-cultured with brain tumors (BO-BTs) own the greatest recapitulation of brain tumors. Tumor invasion and interactions between tumor cells and brain components could be greatly explored in this model. BO-BTs also offer a humanized platform for testing the therapeutic efficacy and side effects on neurons in preclinical trials. We also introduce the BTOs establishment fused with other advanced techniques, such as 3D bioprinting. So far, over 11 brain tumor types of BTOs have been established, especially for glioblastoma. We conclude BTOs could be a reliable model to understand brain tumors and develop targeted therapies.

The PANoptosis-related signature indicates the prognosis and tumor immune infiltration features of gliomas

Background

Gliomas are the most common primary tumors of the central nervous system, with high heterogeneity and highly variable survival rates. Accurate classification and prognostic assessment are key to the selection of treatment strategies. One hallmark of the tumor is resistance to cell death. PANoptosis, a novel mode of programmed cell death, has been frequently reported to be involved in the innate immunity associated with pathogen infection and played an important role in cancers. However, the intrinsic association of PANoptosis with glioma requires deeper investigation.

Methods

The genetics and expression of the 17 reported PANoptosome-related genes were analyzed in glioma. Based on these genes, patients were divided into two subtypes by consensus clustering analysis. After obtaining the differentially expressed genes between clusters, a prognostic model called PANopotic score was constructed after univariate Cox regression, LASSO regression, and multivariate Cox regression. The expression of the 5 genes included in the PANopotic score was also examined by qPCR in our cohort. The prognostic differences, clinical features, TME infiltration status, and immune characteristics between PANoptotic clusters and score groups were compared, some of which even extended to pan-cancer levels.

Results

Gene mutations, CNVs and altered gene expression of PANoptosome-related genes exist in gliomas. Two PANoptotic clusters were significantly different in prognosis, clinical features, immune characteristics, and mutation landscapes. The 5 genes included in the PANopotic score had significantly altered expression in glioma samples in our cohort. The high PANoptotic score group was inclined to show an unfavorable prognosis, lower tumor purity, worse molecular genetic signature, and distinct immune characteristics related to immunotherapy. The PANoptotic score was considered as an independent prognostic factor for glioma and showed superior prognostic assessment efficacy over several reported models. PANopotic score was included in the nomogram constructed for the potential clinical prognostic application. The associations of PANoptotic score with prognostic assessment and tumor immune characteristics were also reflected at the pan-cancer level.

Conclusion

Molecular subtypes of glioma based on PANoptosome-related genes were proposed and PANoptotic score was constructed with different clinical characteristics of anti-tumor immunity. The potential intrinsic association between PANoptosis and glioma subtypes, prognosis, and immunotherapy was revealed.

Modeling brain and neural crest neoplasms with human pluripotent stem cells

Pluripotent stem cells offer unique avenues to study human-specific aspects of disease and are a highly versatile tool in cancer research. Oncogenic processes and developmental programs often share overlapping transcriptomic and epigenetic signatures, which can be reactivated in induced pluripotent stem cells. With the emergence of brain organoids, the ability to recapitulate brain development and structure has vastly improved, making in vitro models more realistic and hence more suitable for biomedical modeling. This review highlights recent research and current challenges in human pluripotent stem cell modeling of brain and neural crest neoplasms, and concludes with a call for more rigorous quality control and for the development of models for rare tumor subtypes.

Receptor tyrosine kinase (RTK) targeting in pediatric high-grade glioma and diffuse midline glioma: Pre-clinical models and precision medicine

Pediatric high-grade glioma (pHGG), including both diffuse midline glioma (DMG) and non-midline tumors, continues to be one of the deadliest oncologic diagnoses (both henceforth referred to as “pHGG”). Targeted therapy options aimed at key oncogenic receptor tyrosine kinase (RTK) drivers using small-molecule RTK inhibitors has been extensively studied, but the absence of proper in vivo modeling that recapitulate pHGG biology has historically been a research challenge. Thankfully, there have been many recent advances in animal modeling, including Cre-inducible transgenic models, as well as intra-uterine electroporation (IUE) models, which closely recapitulate the salient features of human pHGG tumors. Over 20% of pHGG have been found in sequencing studies to have alterations in platelet derived growth factor-alpha (PDGFRA), making growth factor modeling and inhibition via targeted tyrosine kinases a rich vein of interest. With commonly found alterations in other growth factors, including FGFR, EGFR, VEGFR as well as RET, MET, and ALK, it is necessary to model those receptors, as well. Here we review the recent advances in murine modeling and precision targeting of the most important RTKs in their clinical context. We additionally provide a review of current work in the field with several small molecule RTK inhibitors used in pre-clinical or clinical settings for treatment of pHGG.