Discovery Set Research Articles

Altered HCAR3 expression may underlying the blunted niacin responses of the psychiatric disorders and the risk of schizophrenia.

Blunted niacin response (BNR) was an endophenotype of schizophrenia, but the underlying mechanism remains unclarified. The objective of this study was to verify whether genes associated with BNR pathway constitute the genetic basis and the pathological mechanism of BNR phenotypic psychiatric patients. Two independent sample sets consisting of 971 subjects were enrolled in this study. A total of 62 variants were genotyped in the discovery set, then the related variants were verified in the verification set. The published PGC GWAS data were used to validate the associations between the variants and psychiatry disorders. RT-PCR analysis, eQTL data, and Dual-Luciferase Reporter experiment were used to investigate the potential molecular mechanisms of the variants underlying BNR. The results showed that two SNPs, rs56959712 in HCAR2 and rs2454721 in HCAR3 were significantly associated with niacin response. The risk allele T of rs2454721 could affect the niacin responses of psychiatric patients through elevated HCAR3 gene expression. These two genes, especially HCAR3, were significantly associated with the risk of schizophrenia, as identified in this study and verified using the published GWAS data. HCAR3 is a novel schizophrenia susceptibility gene which is significantly associated with blunted niacin response in schizophrenia. In-depth investigation of HCAR3 is of great significance for uncovering the pathogenesis and propose new therapeutic targets for psychiatric disorders, especially for the BNR subgroup patients.

Altered intestinal microbiota enhances adenoid hypertrophy by disrupting the immune balance.

Adenoid hypertrophy (AH) is a common upper respiratory disorder in children. Disturbances of gut microbiota have been implicated in AH. However, the interplay of alteration of gut microbiome and enlarged adenoids remains elusive. 119 AH children and 100 healthy controls were recruited, and microbiome profiling of fecal samples in participants was performed using 16S rRNA gene sequencing. Fecal microbiome transplantation (FMT) was conducted to verify the effects of gut microbiota on immune response in mice. In AH individuals, only a slight decrease of diversity in bacterial community was found, while significant changes of microbial composition were observed between these two groups. Compared with HCs, decreased abundances of Akkermansia, Oscillospiraceae and Eubacterium coprostanoligenes genera and increased abundances of Bacteroides, Faecalibacterium, Ruminococcus gnavus genera were revealed in AH patients. The abundance of Bacteroides remained stable with age in AH children. Notably, a microbial marker panel of 8 OTUs were identified, which discriminated AH from HC individuals with an area under the curve (AUC) of 0.9851 in the discovery set, and verified in the geographically different validation set, achieving an AUC of 0.9782. Furthermore, transfer of mice with fecal microbiota from AH patients dramatically reduced the proportion of Treg subsets within peripheral blood and nasal-associated lymphoid tissue (NALT) and promoted the expansion of Th2 cells in NALT. These findings highlight the effect of the altered gut microbiota in the AH pathogenesis.

High-throughput detection allied with machine learning for precise monitoring of significant serum metabolic changes in Helicobacter pylori infection

High-throughput detection of large-scale samples is the foundation for rapidly accessing massive metabolic data in precision medicine. Machine learning is a powerful tool for uncovering valuable information hidden within massive data. In this work, we achieved the extraction of a single fingerprinting of 1 μL serum within 5 s through a high-throughput detection platform based on functionalized nanoparticles. We quickly obtained over a thousand serum metabolic fingerprintings (SMFs) including those of individuals with Helicobacter pylori (HP) infection. Combining four classical machine learning models and enrichment analysis, we attempted to extract and confirm useful information behind these SMFs. Based on all fingerprint signals, all four models achieved area under the curve (AUC) values of 0.983–1. In particular, orthogonal partial least squares discriminant analysis (OPLS-DA) model obtained value of 1 in both the discovery and validation sets. Fortunately, we identified six significant metabolic features, all of which can greatly contribute to the monitoring of HP infection, with AUC values ranging from 0.906 to 0.985. The combination of these six significant metabolic features can enable the precise monitoring of HP infection in serum, with over 95 % of accuracy, specificity and sensitivity. The OPLS-DA model displayed optimal performance and the corresponding scatter plot visualized the clear distinction between HP and HC. Interestingly, they exhibit a consistent reduction trend compared to healthy controls, prompting us to explore the possible metabolic pathways and potential mechanism. This work demonstrates the potential alliance between high-throughput detection and machine learning, advancing their application in precision medicine.

Investigating the Cell States and Prognostic Impact of Tumor Microenvironment Ecosystems in Classic Hodgkin Lymphoma

Introduction: Classic Hodgkin lymphoma (cHL) is distinguished from other cancers by the minute fraction of Hodgkin/Reed-Sternberg (HRS) cells (1-5%) in the tumor mass. Therefore, the immune-cell rich microenvironment plays a pivotal role in cHL's development, progression, and response to therapy. However, comprehensive characterization of the malignant and immune cell states in cHL has posed significant challenges due to the rarity of HRS cells and the intrinsic heterogeneity of the disease. To address this challenge, we aimed to characterize the immune cell states of cHL using a combination of experimental and bioinformatics approaches. Methods: We identified patients diagnosed with cHL from 2013-19 managed at participating institutions with either formalin-fixed, paraffin-embedded (FFPE, n = 89) or cryopreserved cell suspensions prepared from excisional biopsies (n = 5). FACS purification of cell populations was performed on our cryopreserved tissues, followed by single cell RNA-seq (scRNA-seq) using 10X Genomics. Subsequently, we used CIBERSORTx to construct a customized signature matrix encompassing the gene expression profiles of 11 distinct cell types, including HRS cells. For our FFPE samples, we performed bulk RNA-seq with digital deconvolution using our signature matrix. Next, we combined the data and employed the EcoTyper machine learning framework to discover dominant cell states and ecosystems using deconvolved transcriptomics data. Validation of our identified cell states was performed in the remaining 2 samples with available scRNA-seq data, and a previously published cHL cohort (n = 110, Steidl et al. NEJM 2010). Results: Using our discovery set of scRNA-seq samples (n = 45,405 cells), we identified distinct cell phenotypes including the HRS cells. Using these data, we developed a gene expression signature matrix allowing for the deconvolution of bulk transcriptomes from 89 cHL cases. Among 18 patients with tumor and cell-free somatic tumor mutation data, we observed a positive correlation between the HRS cell proportion and Variant Allele Frequencies (VAFs) from tumor or plasma DNA sequencing ( R = 0.71 and 0.69, respectively). Employing the EcoTyper framework, we discovered 26 unique cell states and three conserved cellular communities or “ecotypes” in the 68 bulk transcriptomes (Figure A). To validate the cell states identified in our EcoTyper model, we employed additional scRNA-seq on two validation samples (n = 31,195 cells) and spatial transcriptomics using Visium 10X (n = 4 tumors). We recovered the three established ecotypes and 26 cell states in these datasets and found a significant spatial autocorrelation of the ecotypes, verifying the biological validity of our EcoTyper model. Finally, we observed Ecotype-3 is associated with adverse progression-free survival in our cohort as well as an independent validation cohort (Figure B) after adjustment for clinical stage in a multivariable model (HR = 8.78, P = 0.02). Conclusions: Our study represents a significant advancement in unraveling the complexities of the cHL cellular landscape. By leveraging innovative methodologies, we have shed light on key cell states and ecotypes within the tumor microenvironment that have prognostic significance, laying the groundwork for potential novel therapeutic avenues. Further validation in additional cohorts may allow ecotypes to risk stratify patients at diagnosis and aid in identification of poor performing subgroups within current low, intermediate, and high-risk groups.

Comparative Assessment of Pose Prediction Accuracy in RNA-Ligand Docking.

Structure-based virtual high-throughput screening is used in early-stage drug discovery. Over the years, docking protocols and scoring functions for protein-ligand complexes have evolved to improve the accuracy in the computation of binding strengths and poses. In the past decade, RNA has also emerged as a target class for new small-molecule drugs. However, most ligand docking programs have been validated and tested for proteins and not RNA. Here, we test the docking power (pose prediction accuracy) of three state-of-the-art docking protocols on 173 RNA-small molecule crystal structures. The programs are AutoDock4 (AD4) and AutoDock Vina (Vina), which were designed for protein targets, and rDock, which was designed for both protein and nucleic acid targets. AD4 performed relatively poorly. For RNA targets for which a crystal structure of a bound ligand used to limit the docking search space is available and for which the goal is to identify new molecules for the same pocket, rDock performs slightly better than Vina, with success rates of 48% and 63%, respectively. However, in the more common type of early-stage drug discovery setting, in which no structure of a ligand-target complex is known and for which a larger search space is defined, rDock performed similarly to Vina, with a low success rate of ∼27%. Vina was found to have bias for ligands with certain physicochemical properties, whereas rDock performs similarly for all ligand properties. Thus, for projects where no ligand-protein structure already exists, Vina and rDock are both applicable. However, the relatively poor performance of all methods relative to protein-target docking illustrates a need for further methods refinement.

CSIG-24. LINEAR GRADING COEFFICIENT REVEALS EMBRYONIC DEVELOPMENT GENE SIGNATURE AS TREATMENT-INDEPENDENT PROGNOSTIC FACTOR IN IDH-MUTANT ASTROCYTOMA

Abstract BACKGROUND Since there is variation in prognosis within grade II/III and grade IV IDH-mutant astrocytomas, we sought to identify features that may aid further tumor grading. METHODS RNAseq was performed on IDH-mutant astrocytomas from the CATNON trial and overlaid with matching genome-wide methylation data. This dataset was extended by multi-omics data of IDH-mutant astrocytomas from the TCGA and GLASS-NL studies. We defined a linear grading coefficient (LGC) based on calibrated DNA methylation-based probability scores and performed differential gene expression (DGE) regression analysis on the LGC. Statistical analyses were intersected across the CATNON and TCGA cohorts ("discovery set") and validated on GLASS-NL. Single-nucleus RNA-sequencing (snRNAseq) data of high-grade IDH-mutant astrocytomas were utilized to explore the source of expression signals. Immunofluorescence was used to confirm findings on protein level. RESULTS DGE regression analysis revealed that higher-grade IDH-mutant tumors (high LGC) exhibited mRNA upregulation of key embryonic developmental transcription factors (HOX[A/C/D], NKX, PAX, and TBX). Although the vast majority of CpG sites in LGC high tumors were hypomethylated, homeobox genes were strongly hypermethylated in our discovery set. HOX gene clusters showed gene-ordered (differential) expression where e.g. HOXD10 >D9 >D8 >D4 >D1 and HOXA3 >A7 >A9 >A13 suggesting derepression of enhancer sequences up/downstream of the gene cluster. Our expression-based embryonic development signature was positively correlated with, but not restricted to, the G-CIMP-low subtype and homozygous CDKN2AB deletion. The signature was prognostic (log-rank test) on both the discovery (CATNON: p< 0.0005, TCGA: p< 0.0005) and validation set (GLASS-NL: p< 0.005). As our signature was developed on molecular data from primary resections, its prognostic value is treatment-independent. SnRNAseq on high-grade IDH-mutant astrocytomas indicated expression of embryonic development genes from tumor cells, not restricted to the stemness program. CONCLUSION Our study highlights upregulation of embryonic development genes as potential molecular mechanism associated with malignancy of IDH-mutant astrocytomas.

Longitudinal modeling of cardiac function to identify trajectories with differential risk of heart failure

Abstract Background Impairments in left ventricular (LV) systolic and diastolic function are common precursors of clinical heart failure (HF) and underlie HF risk. Little data exist regarding trajectories of cardiac function in mid- to late-life, when HF risk is greatest, and the extent to which such trajectories may differentially relate to risk of HF with preserved ejection fraction (HFpEF) or HF with reduced ejection fraction (HFrEF). Purpose To (1) identify multidimensional trajectories of cardiac function from mid- to late-life using serial echocardiographic data, (2) apply the resulting models to predict trajectory membership in a separate sample, and (3) determine associations between assigned trajectory and incident HFpEF and HFrEF. Methods We used a Bayesian nonparametric trajectory approach to jointly model longitudinal measures of LV ejection fraction (LVEF) and E/A ratio among 747 participants in two longitudinal cohort studies (the Atherosclerosis Risk in Communities [ARIC] study and Jackson Heart Study [JHS]) who underwent ≥2 echocardiograms in 2000-04 (mean age 65±5 years), 2011-2013 (age 75±5 years), and 2018-19 (age 81±5 years). Using the trajectories inferred from this modeling, we applied Bayes theorem to assign 4,419 ARIC-only participants to their most likely trajectory using single time-point measures of LVEF and E/A ratio acquired in 2011-2013 (age 75±5 years). These assignments were treated as a categorical variable in subsequent analysis. In ARIC-only participants, we used Cox proportional hazards models adjusted for age, sex, obesity, hypertension, diabetes, chronic kidney disease, and coronary heart disease to assess the relationship between assigned trajectory and incident HF. Results In the discovery set (n=747), we identified six trajectories of cardiac function based jointly on longitudinal changes in LVEF and E/A ratio (Figure 1). The pink (prevalence 50%) and light green (17%) trajectories showed increasing LVEF and decreasing E/A ratio with increasing age, while the red trajectory (22%) demonstrated a lack of increase in LVEF. The dark green trajectory (4%) demonstrated progressive LVEF decline, the orange trajectory (2%) steep LVEF decline and rise in E/A ratio, and the blue trajectory (4%) rising E/A ratio despite increasing in LVEF. Analysis of the ARIC-only participants (n=4,419) showed that trajectory membership was differentially associated with risk of incident HFpEF and HFrEF (Figure 2). Compared to the pink trajectory, membership in the orange trajectory was associated with incident HFpEF and HFrEF, the red and the dark green trajectory associated with HFrEF alone, and the blue trajectory with HFpEF alone. Conclusion(s) These results demonstrate the feasibility of identifying trajectories of cardiac function with multidimensional longitudinal data and may have implications for HF risk prediction.

Circulating Extracellular Vesicle-Derived microRNAs as Novel Diagnostic and Prognostic Biomarkers for Non-Viral-Related Hepatocellular Carcinoma.

Extracellular vesicle-derived microRNAs (EV-miRNAs) are promising circulating biomarkers for chronic liver disease. In this study, we explored the potential significance of plasma EV-miRNAs in non-hepatitis B-, non-hepatitis C-related HCC (NBNC-HCC). We compared, using the NanoString method, plasma EV-miRNA profiles between NBNC-HCC and control groups including patients with non-alcoholic fatty liver disease (NAFLD) and healthy controls. The differentially expressed EV-miRNAs were validated in another set of plasma samples by qRT-PCR. A total of 66 significantly differentially expressed EV-miRNAs between the HCC and the control groups were identified in the discovery set. In the validation cohort, including plasma samples of 70 NBNC-HCC patients, 70 NAFLD patients, and 35 healthy controls, 5 plasma EV-miRNAs were significantly elevated in HCC, which included miR-19-3p, miR-16-5p, miR-223-3p, miR-30d-5p, and miR-451a. These miRNAs were found to participate in several cancer-related signaling pathways based on bioinformatic analysis. Among them, EV-miR-19-3p exhibited the best diagnostic performance and displayed a high sensitivity for detecting alpha-fetoprotein-negative HCC and early-stage HCC. In multivariate analysis, a high EV-miR-19-3p level was demonstrated as an independently unfavorable predictor of overall survival in patients with NBNC-HCC. In conclusion, our data have indicated, for the first time, that EV-miR-19-3p could serve as a novel circulating biomarker for the diagnosis and prognosis of NBNC-HCC.

Subpopulation treatment effect pattern plot analysis: a prognostic model for distant recurrence-free survival to estimate delayed adjuvant chemotherapy initiation effect in triple-negative breast cancer.

Triple-negative breast cancer (TNBC) is a heterogeneous disease associated with a poor prognosis. Delaying in time to start adjuvant chemotherapy (TTC) has been related to an increased risk of distant recurrence-free survival (DRFS). We aimed to develop a prognostic model to estimate the effects of delayed TTC among TNBC risk subgroups. We analyzed 687 TNBC patients who received adjuvant chemotherapy at the Instituto Nacional de Enfermedades Neoplasicas (Lima, Peru). Database was randomly divided to create a discovery set (n=344) and a validation set (n=343). Univariate and multivariate Cox regression models were performed to identify prognostic factors for DRFS. Risk stratification was implemented through two models developed based on proportional hazard ratios from significant clinicopathological characteristics. Subpopulation treatment effect pattern plot (STEPP) analysis was performed to determine the best prognostic cut-off points for stratifying TNBC subgroups according to risk scores and estimate Kaplan-Meier differences in 10-year DRFS comparing TTC (≤30 vs.>30 days). In univariate analysis, patients aged ≥70 years (HR=4.65; 95% CI: 2.32-9.34; p=<0.001), those at stages pT3-T4 (HR=3.28; 95% CI: 1.57-6.83; p=0.002), and pN2-N3 (HR=3.00; 95% CI: 1.90-4.76; p=<0.001) were notably associated with higher risk. STEPP analysis defined three risk subgroups for each model. Model N°01 categorized patients into low (score: 0-31), intermediate (score:32-64), and high-risk (score: 65-100) cohorts; meanwhile, Model N°02: low (score: 0-26), intermediate (score: 27-55), and high (score: 56-100). Kaplan-Meier plots showed that in the discovery set, patients with TTC>30 days experienced a 17.5% decrease in 10-year DRFS rate (95%CI=6.7-28.3), and the impact was more remarkable in patients who belong to the high-risk subgroup (53.3% decrease in 10 years-DRFS rate). Similar results were found in the validation set. We developed two prognostic models based on age, pT, and pN to select the best one to classify TNBC. For Model N°02, delayed adjuvant chemotherapy conferred a higher risk of relapse in patients ≥70 years and who were characterized by pT3/T4 and pN2/N3. Thus, more efforts should be considered to avoid delayed TTC in TNBC patients, especially those in high-risk subgroups.

A large-scale retrospective study enabled deep-learning based pathological assessment of frozen procurement kidney biopsies to predict graft loss and guide organ utilization

Lesion scores on procurement donor biopsies are commonly used to guide organ utilization for deceased-donor kidneys. However, frozen sections present challenges for histological scoring, leading to inter- and intra-observer variability and inappropriate discard. Therefore, we constructed deep-learning based models to recognize kidney tissue compartments in hematoxylin &eosin-stained sections from procurement needle biopsies performed nationwide in years 2011-2020. To do this, we extracted whole-slide abnormality features from 2431 kidneys, and correlated with pathologists’ scores and transplant outcomes. A Kidney Donor Quality Score (KDQS) was derived and used in combination with recipient demographic and peri-transplant characteristics to predict graft loss or assist organ utilization. The performance on wedge biopsies was additionally evaluated. Our model identified 96% and 91% of normal/sclerotic glomeruli respectively; 94% of arteries/arterial intimal fibrosis; 90% of tubules. Whole-slide features of Sclerotic Glomeruli (GS)%, Arterial Intimal Fibrosis (AIF)%, and Interstitial Space Abnormality (ISA)% demonstrated strong correlations with corresponding pathologists’ scores of all 2431 kidneys, but had superior associations with post-transplant estimated glomerular filtration rates in 2033 and graft loss in 1560 kidneys. The combination of KDQS and other factors predicted one- and four-year graft loss in a discovery set of 520 kidneys and a validation set of 1040 kidneys. By using the composite KDQS of 398 discarded kidneys due to “biopsy findings”, we suggest that if transplanted, 110 discarded kidneys could have had similar survival to that of other transplanted kidneys. Thus, our composite KDQS and survival prediction models may facilitate risk stratification and organ utilization while potentially reducing unnecessary organ discard.

Risk stratification of papillary thyroid cancers using multidimensional machine learning.

Papillary thyroid cancer (PTC) is one of the most common endocrine malignancies with different risk levels. However, preoperative risk assessment of PTC is still a challenge in the worldwide. Here, the authors first report a Preoperative Risk Assessment Classifier for PTC (PRAC-PTC) by multidimensional features including clinical indicators, immune indices, genetic feature, and proteomics. The 558 patients collected from June 2013 to November 2020 were allocated to three groups: the discovery set [274 patients, 274 formalin-fixed paraffin-embedded (FFPE)], the retrospective test set (166 patients, 166 FFPE), and the prospective test set (118 patients, 118 fine-needle aspiration). Proteomic profiling was conducted by FFPE and fine-needle aspiration tissues from the patients. Preoperative clinical information and blood immunological indices were collected. The BRAFV600E mutation were detected by the amplification refractory mutation system. The authors developed a machine learning model of 17 variables based on the multidimensional features of 274 PTC patients from a retrospective cohort. The PRAC-PTC achieved areas under the curve (AUC) of 0.925 in the discovery set and was validated externally by blinded analyses in a retrospective cohort of 166 PTC patients (0.787 AUC) and a prospective cohort of 118 PTC patients (0.799 AUC) from two independent clinical centres. Meanwhile, the preoperative predictive risk effectiveness of clinicians was improved with the assistance of PRAC-PTC, and the accuracies reached at 84.4% (95% CI: 82.9-84.4) and 83.5% (95% CI: 82.2-84.2) in the retrospective and prospective test sets, respectively. This study demonstrated that the PRAC-PTC that integrating clinical data, gene mutation information, immune indices, high-throughput proteomics and machine learning technology in multicentre retrospective and prospective clinical cohorts can effectively stratify the preoperative risk of PTC and may decrease unnecessary surgery or overtreatment.

Diverse regulated cell death modes predict the immune microenvironment and drug sensitivity in lung adenocarcinoma.

Integrated action modes of regulated cell death (RCD)in lung adenocarcinoma (LUAD)have not been comprehensively dissected. Here, we adopted 15 RCD modes, including 1350 related genes, and established RCD signature scores. We found that LUAD patients with high RCD scores had a significantly worse prognosis in all four different cohorts (TCGA, KM-plotter, GSE31210, and GSE30219). Our nomogram established based on the RCD score and clinical characteristics performed well in both the discovery and validation sets. There was a close correlation between the RCD scores and LUAD molecular subtypes identified by unsupervised consensus clustering. Furthermore, we profiled the tumor microenvironment via deconvolution and found significant differences in immune activity, transcription factor activity and molecular pathway enrichment between the RCD-high and RCD-low groups. More importantly, we revealed that the regulation of antigen presentation is the crucial mechanism underlying RCD. In addition, higher RCD scores predict poorer sensitivity to multiple therapeutic drugs, which indicates that RCD scores may serve as a promising predictor of chemotherapy and immunotherapy outcomes. In summary, this work is the first to reveal the internal links between RCD modes, LUAD, and cancer immunity and highlights the necessity of RCD scores in personalizing treatment plans.

"Soap bubble" sign as an imaging marker for posterior fossa ependymoma Group B.

To investigate the predictive value of the "soap bubble" sign on molecular subtypes (Group A [PFA] and Group B [PFB]) of posterior fossa ependymomas (PF-EPNs). MRI scans of 227 PF-EPNs (internal retrospective discovery set) were evaluated by two independent neuroradiologists to assess the "soap bubble" sign, which was defined as clusters of cysts of various sizes that look like "soap bubbles" on T2-weighted images. Two independent cohorts (external validation set [n = 31] and prospective validation set [n = 27]) were collected to validate the "soap bubble" sign. Across three datasets, the "soap bubble" sign was observed in 21 PFB cases (7.4% [21/285] of PF-EPNs and 12.9% [21/163] of PFB); none in PFA. Analysis of the internal retrospective discovery set demonstrated substantial interrater agreement (1st Rating: κ = 0.71 [0.53-0.90], 2nd Rating: κ = 0.83 [0.68-0.98]) and intrarater agreement (Rater 1: κ = 0.73 [0.55-0.91], Rater 2: κ = 0.74 [0.55-0.92]) for the "soap bubble" sign; all 13 cases positive for the "soap bubble" sign were PFB (p = 0.002; positive predictive value [PPV] = 100%, negative predictive value [NPV] = 44%, sensitivity = 10%, specificity = 100%). The findings from the external validation set and the prospective validation set were similar, all cases positive for the "soap bubble" sign were PFB (p < 0.001; PPV = 100%). The "soap bubble" sign represents a highly specific imaging marker for the PFB molecular subtype of PF-EPNs.

Genome-wide DNA methylation analysis identifies potent CpG signature for temzolomide response in non-G-CIMP glioblastomas with unmethylated MGMT promoter: MGMT-dependent roles of GPR81.

To identify potent DNA methylation candidates that could predict response to temozolomide (TMZ) in glioblastomas (GBMs) that do not have glioma-CpGs island methylator phenotype (G-CIMP) but have an unmethylated promoter of O-6-methylguanine-DNA methyltransferase (unMGMT). The discovery-validation approach was planned incorporating a series of G-CIMP-/unMGMT GBM cohorts with DNA methylation microarray data and clinical information, to construct multi-CpG prediction models. Different bioinformatic and experimental analyses were performed for biological exploration. By analyzing discovery sets with radiotherapy (RT) plus TMZ versus RT alone, we identified a panel of 64 TMZ efficacy-related CpGs, from which a 10-CpG risk signature was further constructed. Both the 64-CpG panel and the 10-CpG risk signature were validated showing significant correlations with overall survival of G-CIMP-/unMGMT GBMs when treated with RT/TMZ, rather than RT alone. The 10-CpG risk signature was further observed for aiding TMZ choice by distinguishing differential outcomes to RT/TMZ versus RT within each risk subgroup. Functional studies on GPR81, the gene harboring one of the 10 CpGs, indicated its distinct impacts on TMZ resistance in GBM cells, which may be dependent on the status of MGMT expression. The 64 TMZ efficacy-related CpGs and in particular the 10-CpG risk signature may serve as promising predictive biomarker candidates for guiding optimal usage of TMZ in G-CIMP-/unMGMT GBMs.

Identification of a serum-based microRNA signature that detects recurrent oral squamous cell carcinoma before it is clinically evident

BackgroundSurvival rates for oral squamous cell carcinoma (OSCC) have remained poor for decades, a fact largely attributable to late-stage diagnoses and high recurrence rates. We report analysis of serum miRNA expression in samples from patients with high-risk oral lesions (HRL, including OSCC/carcinoma in situ lesions) and healthy non-cancer controls, with the aim of non-invasively detecting primary or recurrent disease before it is clinically evident.MethodsDiscovery, test, and validation sets were defined from a total of 468 serum samples (305 HRL and 163 control samples). Samples were analysed using multiple qRT-PCR platforms.ResultsA two-miRNA classifier comprised of miR-125b-5p and miR-342-3p was defined following discovery and test analyses. Analysis in an independent validation cohort reported sensitivity and specificity of ~74% for this classifier. Significantly, when this classifier was applied to serial serum samples taken from patients both before treatment and during post-treatment surveillance, it identified recurrence an average of 15 months prior to clinical presentation.ConclusionsThese results indicate this serum miRNA classifier is effective as a simple, non-invasive monitoring tool for earlier detection of recurrent disease when lesions are typically smaller and amenable to a wider array of treatment options to improve survival.

Microanalytical Mass Spectrometry with Super-Resolution Microscopy Reveals a Proteome Transition During Development of the Brain's Circadian Pacemaker.

During brain development, neuronal proteomes are regulated in part by changes in spontaneous and sensory-driven activity in immature neural circuits. A longstanding model for studying activity-dependent circuit refinement is the developing mouse visual system where the formation of axonal projections from the eyes to the brain is influenced by spontaneous retinal activity prior to the onset of vision and by visual experience after eye-opening. The precise proteomic changes in retinorecipient targets that occur during this developmental transition are unknown. Here, we developed a microanalytical proteomics pipeline using capillary electrophoresis (CE) electrospray ionization (ESI) mass spectrometry (MS) in the discovery setting to quantify developmental changes in the chief circadian pacemaker, the suprachiasmatic nucleus (SCN), before and after the onset of photoreceptor-dependent visual function. Nesting CE-ESI with trapped ion mobility spectrometry time-of-flight (TOF) mass spectrometry (TimsTOF PRO) doubled the number of identified and quantified proteins compared to the TOF-only control on the same analytical platform. From 10 ng of peptide input, corresponding to <∼0.5% of the total local tissue proteome, technical triplicate analyses identified 1894 proteins and quantified 1066 proteins, including many with important canonical functions in axon guidance, synapse function, glial cell maturation, and extracellular matrix refinement. Label-free quantification revealed differential regulation for 166 proteins over development, with enrichment of axon guidance-associated proteins prior to eye-opening and synapse-associated protein enrichment after eye-opening. Super-resolution imaging of select proteins using STochastic Optical Reconstruction Microscopy (STORM) corroborated the MS results and showed that increased presynaptic protein abundance pre/post eye-opening in the SCN reflects a developmental increase in synapse number, but not presynaptic size or extrasynaptic protein expression. This work marks the first development and systematic application of TimsTOF PRO for CE-ESI-based microproteomics and the first integration of microanalytical CE-ESI TimsTOF PRO with volumetric super-resolution STORM imaging to expand the repertoire of technologies supporting analytical neuroscience.

Spatial Proteome Analysis Identifies Lymphocyte CD44 as a Biomarker Associated with SBRT Resistance in Early-Stage Non-Small Cell Lung Cancer

To discover and validate spatially-resolved protein markers associated with resistance to SBRT in early-stage NSCLC patients. We initially evaluated a discovery cohort of 44 early-stage NSCLC patients treated with SBRT as first-line treatment at the Shandong cancer hospital. Using the GeoMx DSP system, 71 proteins were measured in five molecular compartments (tumor, leukocyte, lymphocyte, macrophage, and stroma) on pre-treatment samples. Candidate biomarkers were orthogonally validated with the Gem AQUA method of quantitative immunofluorescence (QIF). For internal independent cohort validation, we assessed pre-treatment samples derived from 150 NSCLC patients who receive radiotherapy. We further analyzed 100 radiotherapy untreated patients with operable NSCLC to address prognostic significance. Using continuous log-scaled data, we identified CD44 expression in the lymphocyte compartment (CD3+) as a novel predictor of poor progression-free survival (PFS) (multivariate HR = 7.323, p = 0.0079) and overall survival (OS) (multivariate HR = 8.65, p = 0.028) in the discovery set. High CD44 expression in the tumor compartment (pan-cytokeratin, CK+) predicted significantly shorter OS (multivariate HR = 2.208, p = 0.0212), with no significant difference in PFS. We validated by QIF that lymphocyte CD44 levels were associated with resistance to SBRT therapy and prognostic for poor outcomes. Using QIF in an independent radiotherapy treated cohort, we validated that CD44 levels in the lymphocyte compartment were associated with poor PFS and OS. High lymphocyte cell CD44 was not prognostic in non-radiotherapy-treated cohort. Using DSP data, intratumoral regions with elevated lymphocyte cell CD44 expression showed prominent upregulation of CD127, ARG1 and VISTA in the discovery Cohort. In conclusion, we identified and validated lymphocyte cell CD44 as a biomarker indicative of resistance to SBRT or radiotherapy in patients with NSCLC. Further evaluation is warranted to address the predictive value of lymphocyte cell CD44 in multi-institutional studies and clinical trials.

Meta-learning for transformer-based prediction of potent compounds

For many machine learning applications in drug discovery, only limited amounts of training data are available. This typically applies to compound design and activity prediction and often restricts machine learning, especially deep learning. For low-data applications, specialized learning strategies can be considered to limit required training data. Among these is meta-learning that attempts to enable learning in low-data regimes by combining outputs of different models and utilizing meta-data from these predictions. However, in drug discovery settings, meta-learning is still in its infancy. In this study, we have explored meta-learning for the prediction of potent compounds via generative design using transformer models. For different activity classes, meta-learning models were derived to predict highly potent compounds from weakly potent templates in the presence of varying amounts of fine-tuning data and compared to other transformers developed for this task. Meta-learning consistently led to statistically significant improvements in model performance, in particular, when fine-tuning data were limited. Moreover, meta-learning models generated target compounds with higher potency and larger potency differences between templates and targets than other transformers, indicating their potential for low-data compound design.

Integrated study of systemic and local airway transcriptomes in asthma reveals causal mediation of systemic effects by airway key drivers

BackgroundSystemic and local profiles have each been associated with asthma, but parsing causal relationships between system-wide and airway-specific processes can be challenging. We sought to investigate systemic and airway processes in asthma and their causal relationships.MethodsThree hundred forty-one participants with persistent asthma and non-asthmatic controls were recruited and underwent peripheral blood mononuclear cell (PBMC) collection and nasal brushing. Transcriptome-wide RNA sequencing of the PBMC and nasal samples and a series of analyses were then performed using a discovery and independent test set approach at each step to ensure rigor. Analytic steps included differential expression analyses, coexpression and probabilistic causal (Bayesian) network constructions, key driver analyses, and causal mediation models.ResultsAmong the 341 participants, the median age was 13 years (IQR = 10–16), 164 (48%) were female, and 200 (58.7%) had persistent asthma with mean Asthma Control Test (ACT) score 16.6 (SD = 4.2). PBMC genes associated with asthma were enriched in co-expression modules for NK cell-mediated cytotoxicity (fold enrichment = 4.5, FDR = 6.47 × 10−32) and interleukin production (fold enrichment = 2.0, FDR = 1.01 × 10−15). Probabilistic causal network and key driver analyses identified NK cell granule protein (NKG7, fold change = 22.7, FDR = 1.02 × 10−31) and perforin (PRF1, fold change = 14.9, FDR = 1.31 × 10−22) as key drivers predicted to causally regulate PBMC asthma modules. Nasal genes associated with asthma were enriched in the tricarboxylic acid (TCA) cycle module (fold enrichment = 7.5 FDR = 5.09 × 10−107), with network analyses identifying G3BP stress granule assembly factor 1 (G3BP1, fold change = 9.1 FDR = 2.77 × 10−5) and InaD-like protein (INADL, fold change = 5.3 FDR = 2.98 × 10−9) as nasal key drivers. Causal mediation analyses revealed that associations between PBMC key drivers and asthma are causally mediated by nasal key drivers (FDR = 0.0076 to 0.015).ConclusionsIntegrated study of the systemic and airway transcriptomes in a well-phenotyped asthma cohort identified causal key drivers of asthma among PBMC and nasal transcripts. Associations between PBMC key drivers and asthma are causally mediated by nasal key drivers.

Efficient diagnosis of IDH-mutant gliomas: 1p/19qNET assesses 1p/19q codeletion status using weakly-supervised learning

Accurate identification of molecular alterations in gliomas is crucial for their diagnosis and treatment. Although, fluorescence in situ hybridization (FISH) allows for the observation of diverse and heterogeneous alterations, it is inherently time-consuming and challenging due to the limitations of the molecular method. Here, we report the development of 1p/19qNET, an advanced deep-learning network designed to predict fold change values of 1p and 19q chromosomes and classify isocitrate dehydrogenase (IDH)-mutant gliomas from whole-slide images. We trained 1p/19qNET on next-generation sequencing data from a discovery set (DS) of 288 patients and utilized a weakly-supervised approach with slide-level labels to reduce bias and workload. We then performed validation on an independent validation set (IVS) comprising 385 samples from The Cancer Genome Atlas, a comprehensive cancer genomics resource. 1p/19qNET outperformed traditional FISH, achieving R2 values of 0.589 and 0.547 for the 1p and 19q arms, respectively. As an IDH-mutant glioma classifier, 1p/19qNET attained AUCs of 0.930 and 0.837 in the DS and IVS, respectively. The weakly-supervised nature of 1p/19qNET provides explainable heatmaps for the results. This study demonstrates the successful use of deep learning for precise determination of 1p/19q codeletion status and classification of IDH-mutant gliomas as astrocytoma or oligodendroglioma. 1p/19qNET offers comparable results to FISH and provides informative spatial information. This approach has broader applications in tumor classification.