Discovery Dataset Research Articles

Acute hematologic toxicity prediction using dosimetric and radiomics features in patients with cervical cancer: does the treatment regimen matter?

Acute hematologic toxicity (HT) is a prevalent adverse tissue reaction observed in cervical cancer patients undergoing chemoradiotherapy (CRT), which may lead to various negative effects such as compromised therapeutic efficacy and prolonged treatment duration. Accurate prediction of HT occurrence prior to CRT remains challenging. A discovery dataset comprising 478 continuous cervical cancer patients (140 HT patients) and a validation dataset consisting of 205 patients (52 HT patients) were retrospectively enrolled. Both datasets were categorized into the CRT group and radiotherapy (RT)-alone group based on the treatment regimen, i.e., whether chemotherapy was administered within the focused RT duration. Radiomics features were derived by contouring three regions of interest (ROIs)-bone marrow (BM), femoral head (FH), and clinical target volume (CTV)-on the treatment planning CT images before RT. A comprehensive model combining the radiomics features as well as the demographic, clinical, and dosimetric features was constructed to classify patients exhibiting acute HT symptoms in the CRT group, RT group, and combination group. Furthermore, the time-to-event analysis of the discriminative ROI was performed on all patients with acute HT to understand the HT temporal progression. Among three ROIs, BM exhibited the best performance in classifying acute HT, which was verified across all patient groups in both discovery and validation datasets. Among different patient groups in the discovery dataset, acute HT was more precisely predicted in the CRT group [area under the curve (AUC) = 0.779, 95% CI: 0.657-0.874] than that in the RT-alone (AUC = 0.686, 95% CI: 0.529-0.817) or combination group (AUC = 0.748, 95% CI: 0.655-0.827). The predictive results in the validation dataset similarly coincided with those in the discovery dataset: CRT group (AUC = 0.802, 95% CI: 0.669-0.914), RT-alone group (AUC = 0.737, 95% CI: 0.612-0.862), and combination group (AUC = 0.793, 95% CI: 0.713-0.874). In addition, distinct feature sets were adopted for different patient groups. Moreover, the predicted HT risk of BM was also indicative of the HT temporal progression. HT prediction in cervical patients is dependent on both the treatment regimen and ROI selection, and BM is closely related to the occurrence and progression of HT, especially for CRT patients.

Genetic assessment of efficacy and safety profiles of coagulation cascade proteins identifies Factors II and XI as actionable anticoagulant targets.

Anticoagulants are routinely used by millions of patients worldwide to prevent blood clots. Yet, problems with anticoagulant therapy remain, including a persistent and cumulative bleeding risk in patients undergoing prolonged anticoagulation. New safer anticoagulant targets are needed. To prioritize anticoagulant targets with the strongest efficacy [venous thromboembolism (VTE) prevention] and safety (low bleeding risk) profiles, we performed two-sample Mendelian randomization and genetic colocalization. We leveraged three large-scale plasma protein data sets (deCODE as discovery data set and Fenland and Atherosclerosis Risk in Communities as replication data sets] and one liver gene expression data set (Institut Universitaire de Cardiologie et de Pneumologie de Québec bariatric biobank) to evaluate evidence for a causal effect of 26 coagulation cascade proteins on VTE from a new genome-wide association meta-analysis of 44 232 VTE cases and 847 152 controls, stroke subtypes, bleeding outcomes, and parental lifespan as an overall measure of efficacy/safety ratio. A 1 SD genetically predicted reduction in F2 blood levels was associated with lower risk of VTE [odds ratio (OR) = 0.44, 95% confidence interval (CI) = 0.38-0.51, P = 2.6e-28] and cardioembolic stroke risk (OR = 0.55, 95% CI = 0.39-0.76, P = 4.2e-04) but not with bleeding (OR = 1.13, 95% CI = 0.93-1.36, P = 2.2e-01). Genetically predicted F11 reduction was associated with lower risk of VTE (OR = 0.61, 95% CI = 0.58-0.64, P = 4.1e-85) and cardioembolic stroke (OR = 0.77, 95% CI = 0.69-0.86, P = 4.1e-06) but not with bleeding (OR = 1.01, 95% CI = 0.95-1.08, P = 7.5e-01). These Mendelian randomization associations were concordant across the three blood protein data sets and the hepatic gene expression data set as well as colocalization analyses. These results provide strong genetic evidence that F2 and F11 may represent safe and efficacious therapeutic targets to prevent VTE and cardioembolic strokes without substantially increasing bleeding risk.

Risk Score Model for Predicting COVID-19 Progression in Iranian Patients: Development and Validation Study

Background: The recent novel coronavirus disease 2019 (COVID-19) pandemic has underlined the importance of risk score models in public health emergencies. This study aimed to develop a risk prediction score to identify high-risk hospitalized patients for disease progression on admission. Methods: This prospective cohort study included 171 COVID-19 patients, identified through the reverse transcription polymerase chain reaction test, admitted to Bohlool Hospital in Gonabad City, Iran, between April 4 and June 5, 2021. The patients’ demographic, clinical, and laboratory data were collected upon admission, and clinical outcomes were monitored until the end of the study. The discovery dataset (80% of the data) was used to develop the risk score model based on clinical and laboratory features and patient characteristics to predict COVID-19 progression. An additive risk score model was developed based on the regression coefficients of the significant variables in a multiple logistic regression model. The performance of the risk score model was evaluated on the validation dataset (20% of the data) using the receiver operating characteristic (ROC) curve. Statistical analyses were performed with SPSS software, version 21. Results: The Mean±SD for age of participants was 59.54±20.52 years, and 48.6% were male. Most patients (82.5%) fully recovered or showed improvement, while 5.2% experienced disease progression and 12.3% died. Three variables, interleukin-6, neutrophil-to-lymphocyte ratio, and lung involvement, were found to be significant in predicting risk, with a good discriminatory ability, having an area under the ROC curve of 0.970 (95% CI, 0.935%, 1.00%) in the discovery set and 0.973 (95% CI, 0.923%, 1.00%) in the validation set. Conclusion: The developed risk score model in this study can be used as a clinical diagnostic tool to identify COVID-19 patients at higher risk of disease progression and aid in informed decision-making and resource utilization in similar situations, such as respiratory disease outbreaks in the post-corona era.

In search of the ratio of miRNA expression as robust biomarkers for constructing stable diagnostic models among multi-center data.

MicroRNAs (miRNAs) are promising biomarkers for the early detection of disease, and many miRNA-based diagnostic models have been constructed to distinguish patients and healthy individuals. To thoroughly utilize the miRNA-profiling data across different sequencing platforms or multiple centers, the models accounting the batch effects were demanded for the generalization of medical application. We conducted transcription factor (TF)-mediated miRNA-miRNA interaction network analysis and adopted the within-sample expression ratios of miRNA pairs as predictive markers. The ratio of the expression values between each miRNA pair turned out to be stable across multiple data sources. A genetic algorithm-based classifier was constructed to quantify risk scores of the probability of disease and discriminate disease states from normal states in discovery, with a validation dataset for COVID-19, renal cell carcinoma, and lung adenocarcinoma. The predictive models based on the expression ratio of interacting miRNA pairs demonstrated good performances in the discovery and validation datasets, and the classifier may be used accurately for the early detection of disease.

Germline genetic regulation of the colorectal tumor immune microenvironment

ObjectiveTo evaluate the contribution of germline genetics to regulating the briskness and diversity of T cell responses in CRC, we conducted a genome-wide association study to examine the associations between germline genetic variation and quantitative measures of T cell landscapes in 2,876 colorectal tumors from participants in the Molecular Epidemiology of Colorectal Cancer Study (MECC).MethodsGermline DNA samples were genotyped and imputed using genome-wide arrays. Tumor DNA samples were extracted from paraffin blocks, and T cell receptor clonality and abundance were quantified by immunoSEQ (Adaptive Biotechnologies, Seattle, WA). Tumor infiltrating lymphocytes per high powered field (TILs/hpf) were scored by a gastrointestinal pathologist. Regression models were used to evaluate the associations between each variant and the three T-cell features, adjusting for sex, age, genotyping platform, and global ancestry. Three independent datasets were used for replication.ResultsWe identified a SNP (rs4918567) near RBM20 associated with clonality at a genome-wide significant threshold of 5 × 10− 8, with a consistent direction of association in both discovery and replication datasets. Expression quantitative trait (eQTL) analyses and in silico functional annotation for these loci provided insights into potential functional roles, including a statistically significant eQTL between the T allele at rs4918567 and higher expression of ADRA2A (P = 0.012) in healthy colon mucosa.ConclusionsOur study suggests that germline genetic variation is associated with the quantity and diversity of adaptive immune responses in CRC. Further studies are warranted to replicate these findings in additional samples and to investigate functional genomic mechanisms.

Mendelian randomization and transcriptome analysis identified immune-related biomarkers for osteoarthritis.

The immune microenvironment assumes a significant role in the pathogenesis of osteoarthritis (OA). However, the current biomarkers for the diagnosis and treatment of OA are not satisfactory. Our study aims to identify new OA immune-related biomarkers to direct the prevention and treatment of OA using multi-omics data. The discovery dataset integrated the GSE89408 and GSE143514 datasets to identify biomarkers that were significantly associated with the OA immune microenvironment through multiple machine learning methods and weighted gene co-expression network analysis (WGCNA). The identified signature genes were confirmed using two independent validation datasets. We also performed a two-sample mendelian randomization (MR) study to generate causal relationships between biomarkers and OA using OA genome-wide association study (GWAS) summary data (cases n = 24,955, controls n = 378,169). Inverse-variance weighting (IVW) method was used as the main method of causal estimates. Sensitivity analyses were performed to assess the robustness and reliability of the IVW results. Three signature genes (FCER1G, HLA-DMB, and HHLA-DPA1) associated with the OA immune microenvironment were identified as having good diagnostic performances, which can be used as biomarkers. MR results showed increased levels of FCER1G (OR = 1.118, 95% CI 1.031-1.212, P = 0.041), HLA-DMB (OR = 1.057, 95% CI 1.045 -1.069, P = 1.11E-21) and HLA-DPA1 (OR = 1.030, 95% CI 1.005-1.056, P = 0.017) were causally and positively associated with the risk of developing OA. The present study identified the 3 potential immune-related biomarkers for OA, providing new perspectives for the prevention and treatment of OA. The MR study provides genetic support for the causal effects of the 3 biomarkers with OA and may provide new insights into the molecular mechanisms leading to the development of OA.

Abstract LB249: Multi-omics cancer risk assessment of Barrett’s esophagus with DNA methylation sequencing from endoscopic brush samples

Abstract Introduction: Barrett’s esophagus (BE) confers increased cancer risk. Endoscopic surveillance is recommended to distinguish non-dysplastic BE (NDBE), low grade dysplasia (LGD), high grade dysplasia (HGD), or adenocarcinoma (EAC) for endoscopic intervention. However, this paradigm is limited by sampling error, leading to missed/interval EAC. Herein, we aimed to classify these samples more accurately by performing whole genome methylation sequencing (WGMS) and utilizing multiple genomic features on endoscopic brush samples. Materials and methods: Endoscopic brush samples from patients with NDBE (n=18), HGD (18), EAC (18), and normal controls (18) were collected and underwent WGMS (discovery dataset). Sequence reads were processed through internal pipeline to extract CpG methylation and read level methylation ratio (alpha value). Copy number aberration (CNA) detection was performed using both CNVkit and ichorCNA. Differential methylated regions (DMRs) were identified through comparing EAC/HGD with NDBE and normal samples. Targeted methylation sequencing for selected DMRs and shallow WGMS were performed on EAC (41), HGD (26), LGD (23), NDBE (42), and normal controls (37) as validation. Ratio of the reads with alpha value>0.6 for each DMR was used for further analysis. Detected CNA segments were summarized into chromosome arm loss or gain (aneuploidy score or AS representing the number of chromosome arms gained or lost in a sample) and trimmed median absolute deviation (tMAD). Classification models were developed using the discovery set and the best performing one was applied to the validation dataset. Results: Genome-wide DNA methylation largely segregated different groups of samples as diagnosed; however, some unexpected group crossings were observed. Top 200 DMRs were identified between EAC/HGD and NDBE and normal samples. CNA changes were widespread in EAC, very common in HGD but rare in NDBE (17/18 EACs,14/18 HGDs, 5/18 NDBEs with AS>0). With tMAD > 0.015, 16/18 EACs, 8/18 HGDs but none of 18 NDBEs passed the threshold. EAC or HGD samples without detectable CNAs had very low “tumor” fraction as determined by methylation-based cell type deconvolution or ichorCNA. Similar results were found in the validation dataset. A multi-omics (AS, tMAD, estimated tumor fraction, and DMR methylation) model could predict EAC, HGD, or NDBE with 0.81 accuracy through cross validation (10X) and 0.73 in the independent dataset. This increased to 0.98 and 0.90 in separating high risk EAC/HGD from NDBE, respectively. Conclusion: Multi-omics data obtained from single WGMS allow for a comprehensive analysis of BE samples with varying dysplasia levels. Using classification modeling, these characteristics can be effectively used to accurately identify and distinguish between EAC and HGD from NDBE, suggesting their utility as adjuncts to endoscopic surveillance biopsies. Citation Format: Zhifu Sun, Caroline L. Matchett, Seth W. Slettedahl, William R. Taylor, Panwen Wang, Calise Berger, Caryn E. Anderson, Melissa A. Passe, Ramona Lansing, Collin E. Chalmers, Patrick H. Foote, Jeanette E. Eckel Passow, Doug W. Mahoney, John B. Kisiel, Prasad G. Iyer. Multi-omics cancer risk assessment of Barrett’s esophagus with DNA methylation sequencing from endoscopic brush samples [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB249.

Dynamic bTMB combined with residual ctDNA improves survival prediction in locally advanced NSCLC patients with chemoradiotherapy and consolidation immunotherapy

BackgroundLiquid biopsy-based biomarkers, including circulating tumor DNA (ctDNA) and blood tumor mutational burden (bTMB), are recognized as promising predictors of prognoses and responses to immune checkpoint inhibitors (ICIs), despite insufficient sensitivity of single biomarker detection. This research aims to determine whether the combinatorial utility of longitudinal ctDNA with bTMB analysis could improve the prognostic and predictive effects. MethodsThis prospective two-center cohort trial, consisting of discovery and validation datasets, enrolled unresectable locally advanced non-small-cell lung cancer (LA-NSCLC) patients and assigned them to chemoradiotherapy (CRT) or CRT + consolidation ICI cohorts from 2018 to 2022. Blood specimens were collected pretreatment, 4 weeks post-CRT, and at progression to assess bTMB and ctDNA using 486-gene next-generation sequencing. Dynamic ∆bTMB was calculated as post-CRT bTMB minus baseline bTMB levels. Decision curve analyses were performed to calculate Concordance index (C-index). ResultsOne hundred twenty-eight patients were enrolled. In the discovery dataset (n = 73), patients treated with CRT and consolidation ICI had significantly longer overall survival (OS; median not reached [NR] vs 20.2 months; P < 0.001) and progression-free survival (PFS; median 25.2 vs 11.4 months; P = 0.011) than those without ICI. Longitudinal analysis demonstrated a significant decrease in ctDNA abundance post-CRT (P < 0.001) but a relative increase with disease progression. Post-CRT detectable residual ctDNA correlated with significantly shorter OS (median 18.3 months vs NR; P = 0.001) and PFS (median 7.3 vs 25.2 months; P < 0.001). For patients with residual ctDNA, consolidation ICI brought significantly greater OS (median NR vs 14.8 months; P = 0.005) and PFS (median 13.8 vs 6.2 months; P = 0.028) benefit, but no significant difference for patients with ctDNA clearance. Dynamic ∆bTMB was predictive of prognosis. Patients with residual ctDNA and increased ∆bTMB (∆bTMB > 0) had significantly worse OS (median 9.0 vs 23.0 months vs NR; P < 0.001) and PFS (median 3.4 vs 7.3 vs 25.2 months; P < 0.001). The combinatorial model integrating post-CRT ctDNA with ∆bTMB had optimal predictive effects on OS (C-index = 0.723) and PFS (C-index = 0.693), outperforming individual features. In the independent validation set, we confirmed residual ctDNA predicted poorer PFS (median 50.8 vs 14.3 months; P = 0.026) but identified more consolidation ICI benefit (median NR vs 8.3 months; P = 0.039). The combined model exhibited a stable predictive advantage (C-index = 0.742 for PFS). ConclusionsThe multiparameter assay integrating qualitative residual ctDNA testing with quantitative ∆bTMB dynamics improves patient prognostic risk stratification and efficacy predictions, allowing for personalized consolidation therapy for LA-NSCLC.

Abstract 4428: Integrated multi-omics analysis reveals systemic and localized metabolic disruptions in colorectal cancer

Abstract Despite technological advances in molecular medicine over the last 30 years, no single approach has proven to be sufficient to meet the diagnostic needs in cancer care. Here we use a multiomic approach leveraging proteomic, metabolomic, and transcriptomic technologies to study the metabolic shifts that occur during colorectal cancer (CRC) progression, Our integrated analysis identified systemic changes to beta oxidation pathways and localized changes to tyrosine metabolism within the tumor, creating a mechanistic, actionable description of the core CRC metabolic program. We conducted untargeted proteomics and metabolomics profiling on serum samples from CRC patients (n=10) and healthy controls (n=10). Our serum proteomics data is generated through sample enrichment on the Seer Proteograph system followed by LC/MS analysis with the Thermo Orbitrap Astral. Our metabolomics approach employs LC/MS assays for unbiased profiling of the serum metabolome, exposome, and lipidome. We supplemented this discovery work with both targeted serum and tumor-specific approaches including targeted proteomics to quantify select targets, inflammatory profiling with Alamar proteomics, public transcriptomics data from TCGA, and in vitro metabolic flux data. These large and diverse datasets were then integrated through joint pathway analysis and network integration. The global serum proteomics and metabolomics profiles generated show distinct separation between healthy and diseased individuals. Joint pathway analysis of these discovery datasets highlighted enrichment in beta oxidation pathways and tyrosine metabolism, among others. Interestingly, the Alamar inflammatory panel revealed that many of the analytes in the tyrosine metabolism pathway were well correlated with inflammatory status, while the beta oxidation signature had a lower correlation. To determine the relationship of these systemic findings from serum to tumor metabolism itself, we integrated tumor-specific transcriptomics data and found alterations to tyrosine metabolism with differences in tyrosine aminotransferase expression. The beta oxidation related genes, on the other hand, were not concordant with the serum findings. However, our in vitro metabolic flux studies have shown beta oxidation in CRC cells is upregulated to provide additional fuel for oxidative phosphorylation. This result suggests that beta oxidation may not be transcriptionally regulated in the tumor but rather a consequence of organismal metabolic rewiring. The results of this work, which is currently expanding into a larger and more diverse patient population, underscores the power of multiomic profiling for enhancing our understanding of metabolic dysregulation in CRC. Ultimately, a comprehensive model of the molecular alterations in CRC will yield a better understanding of tumor phenotypes and inform better diagnostics and therapies. Citation Format: Gary Patti, Ethan Stancliffe, Adam Richardson, Ashima Mehta, Monil Gandhi, Kevin Cho. Integrated multi-omics analysis reveals systemic and localized metabolic disruptions in colorectal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4428.

Causal effects of sleep traits on metabolic syndrome and its components: a Mendelian randomization study.

Previous observational studies have suggested an association between sleep disturbance and metabolic syndrome (MetS). However, it remains unclear whether this association is causal. This study aims to investigate the causal effects of sleep-related traits on MetS using Mendelian randomization (MR). Single-nucleotide polymorphisms strongly associated with daytime napping, insomnia, chronotype, short sleep, and long sleep were selected as genetic instruments from the corresponding genome-wide association studies (GWAS). Summary-level data for MetS were obtained from two independent GWAS datasets. Univariable and multivariable MR analyses were conducted to investigate and verify the causal effects of sleep traits on MetS. The univariable MR analysis demonstrated that genetically predicted daytime napping and insomnia were associated with increased risk of MetS in both discovery dataset (OR daytime napping = 1.630, 95% CI 1.273, 2.086; OR insomnia = 1.155, 95% CI 1.108, 1.204) and replication dataset (OR daytime napping = 1.325, 95% CI 1.131, 1.551; OR insomnia = 1.072, 95% CI 1.046, 1.099). For components, daytime napping was positively associated with triglycerides (beta = 0.383, 95% CI 0.160, 0.607) and waist circumference (beta = 0.383, 95% CI 0.184, 0.583). Insomnia was positively associated with hypertension (OR = 1.101, 95% CI 1.042, 1.162) and waist circumference (beta = 0.067, 95% CI 0.031, 0.104). The multivariable MR analysis indicated that the adverse effect of daytime napping and insomnia on MetS persisted after adjusting for BMI, smoking, drinking, and another sleep trait. Our study supported daytime napping and insomnia were potential causal factors for MetS characterized by central obesity, hypertension, or elevated triglycerides.

Enabling large-scale screening of Barrett’s esophagus using weakly supervised deep learning in histopathology

Timely detection of Barrett’s esophagus, the pre-malignant condition of esophageal adenocarcinoma, can improve patient survival rates. The Cytosponge-TFF3 test, a non-endoscopic minimally invasive procedure, has been used for diagnosing intestinal metaplasia in Barrett’s. However, it depends on pathologist’s assessment of two slides stained with H&E and the immunohistochemical biomarker TFF3. This resource-intensive clinical workflow limits large-scale screening in the at-risk population. To improve screening capacity, we propose a deep learning approach for detecting Barrett’s from routinely stained H&E slides. The approach solely relies on diagnostic labels, eliminating the need for expensive localized expert annotations. We train and independently validate our approach on two clinical trial datasets, totaling 1866 patients. We achieve 91.4% and 87.3% AUROCs on discovery and external test datasets for the H&E model, comparable to the TFF3 model. Our proposed semi-automated clinical workflow can reduce pathologists’ workload to 48% without sacrificing diagnostic performance, enabling pathologists to prioritize high risk cases.

Associations of 50 modifiable risk factors with atrial fibrillation using Mendelian randomization analysis.

Substantial focus has been placed on atrial fibrillation (AF) treatment and associated stroke prevention rather than preventing AF itself. We employed Mendelian randomization (MR) approach to examine the causal relationships between 50 modifiable risk factors (RFs) and AF. Instrumental variables for genetically predicted exposures were derived from corresponding genome-wide association studies (GWASs). Summary-level statistical data for AF were obtained from a GWAS meta-analysis (discovery dataset, N = 1,030,836) and FinnGen (validation dataset, N = 208,594). Univariable and multivariable MR analyses were performed, primarily using inverse variance weighted method with a series of robust sensitivity analyses. Genetic predisposition to insomnia, daytime naps, apnea, smoking initiation, moderate to vigorous physical activity and obesity traits, including body mass index, waist-hip ratio, central and peripheral fat/fat-free mass, exhibited significant associations with an increased risk of AF. Coffee consumption and ApoB had suggestive increased risks. Hypertension (odds ratio (OR) 95% confidence interval (CI): 5.26 (4.42, 6.24)), heart failure (HF) (OR 95% CI, 4.77 (2.43, 9.37)) and coronary artery disease (CAD) (OR 95% CI: 1.20 (1.16, 1.24)) were strongly associated with AF, while college degree, higher education attachment and HDL levels were associated with a decreased AF risk. Reverse MR found a bidirectional relationship between genetically predicted AF and CAD, HF and ischemic stroke. Multivariable analysis further indicated that obesity-related traits, systolic blood pressure and lower HDL levels independently contributed to the development of AF. This study identified several lifestyles and cardiometabolic factors that might be causally related to AF, underscoring the importance of a holistic approach to AF management and prevention.

Transfer learning with graph neural networks for improved molecular property prediction in the multi-fidelity setting

We investigate the potential of graph neural networks for transfer learning and improving molecular property prediction on sparse and expensive to acquire high-fidelity data by leveraging low-fidelity measurements as an inexpensive proxy for a targeted property of interest. This problem arises in discovery processes that rely on screening funnels for trading off the overall costs against throughput and accuracy. Typically, individual stages in these processes are loosely connected and each one generates data at different scale and fidelity. We consider this setup holistically and demonstrate empirically that existing transfer learning techniques for graph neural networks are generally unable to harness the information from multi-fidelity cascades. Here, we propose several effective transfer learning strategies and study them in transductive and inductive settings. Our analysis involves a collection of more than 28 million unique experimental protein-ligand interactions across 37 targets from drug discovery by high-throughput screening and 12 quantum properties from the dataset QMugs. The results indicate that transfer learning can improve the performance on sparse tasks by up to eight times while using an order of magnitude less high-fidelity training data. Moreover, the proposed methods consistently outperform existing transfer learning strategies for graph-structured data on drug discovery and quantum mechanics datasets.

Network Localization of State and Trait of Auditory Verbal Hallucinations in Schizophrenia.

Neuroimaging studies investigating the neural substrates of auditory verbal hallucinations (AVH) in schizophrenia have yielded mixed results, which may be reconciled by network localization. We sought to examine whether AVH-state and AVH-trait brain alterations in schizophrenia localize to common or distinct networks. We initially identified AVH-state and AVH-trait brain alterations in schizophrenia reported in 48 previous studies. By integrating these affected brain locations with large-scale discovery and validation resting-state functional magnetic resonance imaging datasets, we then leveraged novel functional connectivity network mapping to construct AVH-state and AVH-trait dysfunctional networks. The neuroanatomically heterogeneous AVH-state and AVH-trait brain alterations in schizophrenia localized to distinct and specific networks. The AVH-state dysfunctional network comprised a broadly distributed set of brain regions mainly involving the auditory, salience, basal ganglia, language, and sensorimotor networks. Contrastingly, the AVH-trait dysfunctional network manifested as a pattern of circumscribed brain regions principally implicating the caudate and inferior frontal gyrus. Additionally, the AVH-state dysfunctional network aligned with the neuromodulation targets for effective treatment of AVH, indicating possible clinical relevance. Apart from unifying the seemingly irreproducible neuroimaging results across prior AVH studies, our findings suggest different neural mechanisms underlying AVH state and trait in schizophrenia from a network perspective and more broadly may inform future neuromodulation treatment for AVH.

Enhancement of Prediction Model for Students’ Performance in Intelligent Tutoring System

Adapting Artificial Intelligence to Intelligent Tutoring System (ITS) has made teaching and learning more effective. Prediction of students’ performance has gained more interest among researchers to know whether the students master their learning before moving to another topic. For the research scope, we have analyzed numerous Bayesian Knowledge Tracing (BKT) variations in methodology and found that the most precise way to forecast students’ success is through Individualized Bayesian Knowledge Tracing (iBKT). Although iBKT makes a good prediction, iBKT does not consider other knowledge-related elements, such as learning and guess rate, and only uses students’ prior knowledge as the parameters. Due to issues concerning uncertainties in students’ interactions, this study proposes to enhance the prediction function of the iBKT using a feature relating to students’ confidence levels. Thus, this new confidence parameter is defined as P(C), assumed to improve prediction accuracy when forecasting student achievement. The prediction accuracy is tested using the attributes of the ASSISTment and Knowledge Discovery and Data Mining (KDD) datasets as input. In addition, Root Mean Square Error (RMSE) is applied to calculate the performance accuracy of iBKT and enhanced iBKT with the confidence parameter. As a result, the RMSE performance accuracy of iBKT with the confidence parameter shows a low RMSE score for both datasets. The ASSISTment dataset provides a higher prediction when applying the confidence parameter, 0.21190. Therefore, it is concluded that enhancing the confidence parameter is an effective method with accuracy improvement for predicting students’ success in ITS.

The causal effect of two occupational factors on osteoarthritis and rheumatoid arthritis: a Mendelian randomization study.

Osteoarthritis (OA) and rheumatoid arthritis (RA) are two common types of arthritis. We conducted a two-sample Mendelian randomization (MR) study to estimate the causal effects of two common occupational factors-job involves heavy manual or physical work and job involves mainly walking or standing-on OA and RA in individuals of European ancestry. Instruments were chosen from genome-wide association studies (GWASs) that identified independent single nucleotide polymorphisms (SNPs) robustly linked to job involves heavy manual or physical work (N = 263,615) as well as job involves mainly walking or standing (N = 263,556). Summary statistics for OA and RA were taken from the Integrative Epidemiology Unit (IEU) GWAS database; both discovery and replication GWAS datasets were considered. The primary analysis utilized the inverse variance weighted (IVW) MR method supplemented by various sensitivity MR analyses. In the IVW model, we found that genetically predicted job involves heavy manual or physical work was significantly associated with OA in both the discovery [odds ratio (OR) = 1.034, 95% confidence interval (CI): 1.016-1.053, P = 2.257 × 10-4] and replication (OR = 1.857, 95% CI: 1.223-2.822, P = 0.004) analyses. The causal associations were supported in diverse sensitivity analyses. MR analyses suggested no causal effect of genetically predicted job involves heavy manual or physical work on RA. Similarly, our data provided no evidence that genetically predicted job involves mainly walking or standing was related to OA and RA. Our MR study suggests that job involves heavy manual or physical work is a risk factor for OA. It is of utmost importance to create preventive strategies aimed at reducing its impact on OA at such work sites.

Genome-wide association study implicates the role of TBXAS1 in the pathogenesis of depressive symptoms among the Korean population

Although depression is an emerging disorder affecting many people worldwide, most genetic studies have been performed in European descent populations. Herein, a genome-wide association study (GWAS) was conducted in Korean population to elucidate the genomic loci associated with depressive symptoms. Two independent cohorts were used as discovery datasets, which consisted of 6474 (1484 cases and 4990 controls) and 1654 (557 cases and 1097 controls) Korean participants, respectively. The participants were divided into case and control groups based on the Beck Depression Inventory (BDI). Meta-analysis using the two cohorts revealed that rs6945590 was significantly associated with the risk of depressive symptoms [P = 2.83 × 10−8; odds ratio (OR) = 1.23; 95% confidence interval (CI): 1.15–1.33]. This association was validated in other independent cohorts which were another Korean cohort (258 cases and 1757 controls) and the East Asian study of the Psychiatric Genomics Consortium (PGC) (12,455 cases and 85,548 controls). The predicted expression levels of thromboxane A synthase 1 gene (TBXAS1), which encodes the enzyme thromboxane A synthase 1 and participates in the arachidonic acid (AA) cascade, was significantly decreased in the whole blood tissues of the participants with depressive symptoms. Furthermore, Mendelian randomization (MR) analysis showed a causal association between TBXAS1 expression and the risk of depressive symptoms. In conclusion, as the number of risk alleles (A) of rs6945590 increased, TBXAS1 expression decreased, which subsequently caused an increase in the risk of depressive symptoms.

ZNF326 as a potential prognostic and predictive biomarker in stage II colorectal cancer.

Adjuvant chemotherapy is considered for stage II colorectal cancer (CRC) patients with poor prognostic risk factors. However, current stratification algorithms are still insufficient to identify high-risk patients. We conducted a screening strategy to define ZNF326 based on quantitative proteomics in 11 paired CRC patients selected by a nested case-control design, and tested the association between ZNF326 expression level with the prognosis of stage II CRC patients and the benefit from adjuvant chemotherapy in public datasets; further investigation was conducted through subgroup analyses. We found that low ZNF326 expression was significantly associated with a lower 5-year overall survival (OS) rate among stage II patients in both the discovery [P=0.008; hazard ratio (HR): 3.13, 95% confidence interval (CI): 1.29-7.58] and validation (P=0.025; HR: 1.98, 95% CI: 1.08-3.65) cohorts. In the Cox multivariable analysis, low ZNF326 expression was both associated with shorter OS after adjustment for age, sex, and adjuvant chemotherapy in the discovery and validation data sets. Subgroup analyses yielded largely similar results. In a pooled database, the rate of 5-year OS was higher among stage II ZNF326-high tumors who were treated with adjuvant chemotherapy than it was among those who were not treated with adjuvant chemotherapy (P=0.011; HR: 0.28, 95% CI: 0.10-0.80). ZNF326 has the potential to be used in clinical practice for risk classification. ZNF326-low expression level identified a subgroup of patients with high-risk stage II CRC who appeared to less benefit from adjuvant chemotherapy.

ADGRE5-centered Tsurv model in T cells recognizes responders to neoadjuvant cancer immunotherapy.

Neoadjuvant immunotherapy with anti-programmed death-1 (neo-antiPD1) has revolutionized perioperative methods for improvement of overall survival (OS), while approaches for major pathologic response patients' (MPR) recognition along with methods for overcoming non-MPR resistance are still in urgent need. We utilized and integrated publicly-available immune checkpoint inhibitors regimens (ICIs) single-cell (sc) data as the discovery datasets, and innovatively developed a cell-communication analysis pipeline, along with a VIPER-based-SCENIC process, to thoroughly dissect MPR-responding subsets. Besides, we further employed our own non-small cell lung cancer (NSCLC) ICIs cohort's sc data for validation in-silico. Afterward, we resorted to ICIs-resistant murine models developed by us with multimodal investigation, including bulk-RNA-sequencing, Chip-sequencing and high-dimensional cytometry by time of flight (CYTOF) to consolidate our findings in-vivo. To comprehensively explore mechanisms, we adopted 3D ex-vivo hydrogel models for analysis. Furthermore, we constructed an ADGRE5-centered Tsurv model from our discovery dataset by machine learning (ML) algorithms for a wide range of tumor types (NSCLC, melanoma, urothelial cancer, etc.) and verified it in peripheral blood mononuclear cells (PBMCs) sc datasets. Through a meta-analysis of multimodal sequential sc sequencing data from pre-ICIs and post-ICIs, we identified an MPR-expanding T cells meta-cluster (MPR-E) in the tumor microenvironment (TME), characterized by a stem-like CD8+ T cluster (survT) with STAT5-ADGRE5 axis enhancement compared to non-MPR or pre-ICIs TME. Through multi-omics analysis of murine TME, we further confirmed the existence of survT with silenced function and immune checkpoints (ICs) in MPR-E. After verification of the STAT5-ADGRE5 axis of survT in independent ICIs cohorts, an ADGRE5-centered Tsurv model was then developed through ML for identification of MPR patients pre-ICIs and post-ICIs, both in TME and PBMCs, which was further verified in pan-cancer immunotherapy cohorts. Mechanistically, we unveiled ICIs stimulated ADGRE5 upregulation in a STAT5-IL32 dependent manner in a 3D ex-vivo system (3D-HYGTIC) developed by us previously, which marked Tsurv with better survival flexibility, enhanced stemness and potential cytotoxicity within TME. Our research provides insights into mechanisms underlying MPR in neo-antiPD1 and a well-performed model for the identification of non-MPR.

Brain Structural and Functional Damage Network Localization of Suicide

BackgroundExtensive neuroimaging research on brain structural and functional correlates of suicide has produced inconsistent results. Despite increasing recognition that damage in multiple different brain locations that causes the same symptom can map to a common brain network, there is still a paucity of research investigating network localization of suicide. MethodsTo clarify this issue, we initially identified brain structural and functional damage locations in relation to suicide from 63 published studies with 2135 suicidal and 2606 nonsuicidal individuals. By applying novel functional connectivity network mapping to large-scale discovery and validation resting-state functional magnetic resonance imaging datasets, we mapped these affected brain locations to 3 suicide brain damage networks corresponding to different imaging modalities. ResultsThe suicide gray matter volume damage network comprised widely distributed brain areas primarily involving the dorsal default mode, basal ganglia, and anterior salience networks. The suicide task-induced activation damage network was similar to but less extensive than the gray matter volume damage network, predominantly implicating the same canonical networks. The suicide resting-state activity damage network manifested as a localized set of brain regions encompassing the orbitofrontal cortex and middle cingulate cortex. ConclusionsOur findings not only may help reconcile prior heterogeneous neuroimaging results, but also may provide insights into the neurobiological mechanisms of suicide from a network perspective, which may ultimately inform more targeted and effective strategies to prevent suicide.