Expression Data Research Articles

Bayesian Optimal Two-Sample Tests for High-Dimensional Gaussian Populations

We propose minimax optimal Bayesian two-sample tests for testing equality of high-dimensional mean vectors and covariance matrices between two populations. In many applications including genomics and medical imaging, it is natural to assume that only a few entries of two mean vectors or covariance matrices are different. Many existing tests that rely on aggregating the difference between empirical means or covariance matrices are not optimal or yield low power under such setups. Motivated by this, we develop Bayesian two-sample tests employing a divide-and-conquer idea, which is powerful especially when the differences between two populations are rare but large. The proposed two-sample tests manifest closed forms of Bayes factors and allow scalable computations even in high-dimensions. We prove that the proposed tests are consistent under relatively mild conditions compared to existing tests in the literature. Furthermore, the testable regions from the proposed tests turn out to be minimax optimal in terms of rates. Simulation studies show clear advantages of the proposed tests over other state-of-the-art methods in various scenarios. Our tests are also applied to the analysis of the gene expression data of two cancer data sets.

The Biological Function of POLA2 in Hepatocellular Carcinoma.

The role and prognostic value of POLA2 in liver cancer were comprehensively analyzed through TCGA, GEO, and ICGC databases, and the role of POLA2 in liver cancer cells and the regulatory mechanism involved were further verified through cell experiments. Hepatocellular carcinoma (HCC) is the most prevalent malignancy with high morbidity and mortality. Consequently, it is critical to identify robust and reliable predictive biomarkers and therapeutic targets for HCC patients. POLA2 is involved in the regulation of various tumors, but the specific role of POLA2 in HCC has not been reported. The regulatory role and prognostic value of POLA2 in HCC were determined by bioinformatics techniques and cell experiments. The specific role and prognostic value of POLA2 in HCC were comprehensively analyzed by combining the expression data of POLA2 in TCGA, GEO, and ICGC databases and clinical data. In clinical samples, the expression of POLA2 in liver cancer was verified by QPCR. Further, the regulatory role of POLA2 in HCC was explored through cell experiments such as CCK-8, clonal formation experiment, EDU cell proliferation experiment, and flow cytometry. In terms of mechanism exploration, western blot was used to verify the specific regulatory mechanism that POLA2 participated in. Finally, the relationship between POLA2 and immune invasion of HCC was analyzed by using the TIMER database. A POLA2 expression and prognosis analysis of HCC patients was conducted using the TCGA, GEO, and ICGC databases. We hypothesized that POLA2 might be one of the key factors contributing to the HCC progression. According to a combined analysis of TCGA, ICGC, and GEO databases, POLA2 was highly expressed in HCC. This was further confirmed in clinical samples using the qPCR. POLA2 knockdown was also performed in vitro on HCC cell lines to study the changes in their biological behavior. We confirmed that POLA2 was associated with HCC proliferation by CCK-8, Colony Formation, and EDU assay. We verified the POLA2's involvement in cell cycle regulation using flow techniques. The relationship between POLA2 and PI3K/AKT/mTOR pathway was explored using Western Blotting experiments regarding its mechanism. Further analysis revealed that the POLA2 expression was significantly associated with HCC immune infiltration. Our study demonstrated POLA2's importance in HCC development and progression and its potential role as a biomarker for disease progression on multiple levels. POLA2 has an important role in regulating the cell cycle and cell proliferation. By interfering with the cell cycle and proliferation, HCC cell growth is inhibited. Furthermore, POLA2 expression was significantly associated with immune infiltration. POLA2 may play a role in HCC immunotherapy based on its correlation with several immune cell types' genetic markers. The findings of this study are expected to lead to new anticancer strategies for HCC.

Comprehensive Analyses and Experiments Confirmed IGFBP5 as a Prognostic Predictor Based on an Aging-genomic Landscape Analysis of Ovarian Cancer.

Ovarian cancer (OC) is one of the malignant diseases of the reproductive system in elderly women. Aging-related genes (ARGs) were involved in tumor malignancy and cellular senescence, but the specifics of these mechanisms in OC remain unknown. ARGs expression and survival data of OC patients were collected from TCGA and CPTAC databases. Subtype classification was used to identify the roles of hub ARGs in OC progression, including function enrichment, immune infiltration, and drug sensitivity. LASSO regression was utilized to confirm the prognosis significance for these hub ARGs. MTT, EdU, Transwell, and wounding healing analysis confirmed the effect of IGFBP5 on the proliferation and migration ability of OC cells. ARGs were ectopically expressed in OC tissues compared to normal ovary tissues. Three molecular subtypes were divided by ARGs for OC patients. There were significant differences in ferroptosis, m6A methylation, prognosis, immune infiltration, angiogenesis, differentiation level, and drug sensitivity among the three groups. LASSO regression indicated that 4 signatures, FOXO4, IGFBP5, OGG1 and TYMS, had important prognosis significance. Moreover, IGFBP5 was significantly correlated with immune infiltration. The hub ARG, IGFBP5, expression was significantly decreased in OC patients compared to normal women. IGFBP5 could also reduce the migration and proliferation ability of OC cells compared to vector and NC groups. IGFBP5 was correlated with OC prognosis and associated with OC migration and proliferation. This gene may serve as potential prognostic biomarkers and therapeutic targets for OC patients.

Ras-dependent activation of BMAL2 regulates hypoxic metabolism in pancreatic cancer.

To identify drivers of malignancy in human pancreatic ductal adenocarcinoma (PDAC), we performed regulatory network analysis on a large collection of expression profiles from laser capture microdissected samples of PDAC and benign precursors. We discovered that BMAL2 plays a role in the initiation, progression, post resection survival, and KRAS activity in PDAC. Functional analysis of BMAL2 target genes led us to hypothesize that it plays a role in regulating the response to hypoxia, a critical but poorly understood feature of PDAC physiology. Knockout of BMAL2 in multiple human PDAC cell lines revealed effects on viability and invasion, particularly under hypoxic conditions. Loss of BMAL2 also affected glycolysis and other metabolic processes. We found that BMAL2 directly regulates hypoxia-responsive target genes. We also found that BMAL2 is necessary for the stabilization of HIF1A upon exposure to hypoxia, but destabilizes HIF2A under hypoxia. These data demonstrate that BMAL2 is a master transcriptional regulator of hypoxia responses in PDAC and may serve as a long-sought molecular switch that distinguishes HIF1A- and HIF2A-dependent modes of hypoxic metabolism.

Alzheimer DataLENS: An Open Data Analytics Portal for Alzheimer's Disease Research.

Recent Alzheimer's disease (AD) discoveries are increasingly based on studies from a variety of omics technologies on large cohorts. Currently, there is no easily accessible resource for neuroscientists to browse, query, and visualize these complex datasets in a harmonized manner. Create an online portal of public omics datasets for AD research. We developed Alzheimer DataLENS, a web-based portal, using the R Shiny platform to query and visualize publicly available transcriptomics and genetics studies of AD on human cohorts. To ensure consistent representation of AD findings, all datasets were processed through a uniform bioinformatics pipeline. Alzheimer DataLENS currently houses 2 single-nucleus RNA sequencing datasets, over 30 bulk RNA sequencing datasets from 19 brain regions and 3 cohorts, and 2 genome-wide association studies (GWAS). Available visualizations for single-nucleus data include bubble plots, heatmaps, and UMAP plots; for bulk expression data include box plots and heatmaps; for pathways include protein-protein interaction network plots; and for GWAS results include Manhattan plots. Alzheimer DataLENS also links to two other knowledge resources: the AD Progression Atlas and the Astrocyte Atlas. Alzheimer DataLENS is a valuable resource for investigators to quickly and systematically explore omics datasets and is freely accessible at https://alzdatalens.partners.org.

Mitochondria-Related Candidate Genes and Diagnostic Model to Predict Late-Onset Alzheimer's Disease and Mild Cognitive Impairment.

Late-onset Alzheimer's disease (LOAD) is the most common type of dementia, but its pathogenesis remains unclear, and there is a lack of simple and convenient early diagnostic markers to predict the occurrence. Our study aimed to identify diagnostic candidate genes to predict LOAD by machine learning methods. Three publicly available datasets from the Gene Expression Omnibus (GEO) database containing peripheral blood gene expression data for LOAD, mild cognitive impairment (MCI), and controls (CN) were downloaded. Differential expression analysis, the least absolute shrinkage and selection operator (LASSO), and support vector machine recursive feature elimination (SVM-RFE) were used to identify LOAD diagnostic candidate genes. These candidate genes were then validated in the validation group and clinical samples, and a LOAD prediction model was established. LASSO and SVM-RFE analyses identified 3 mitochondria-related genes (MRGs) as candidate genes, including NDUFA1, NDUFS5, and NDUFB3. In the verification of 3 MRGs, the AUC values showed that NDUFA1, NDUFS5 had better predictability. We also verified the candidate MRGs in MCI groups, the AUC values showed good performance. We then used NDUFA1, NDUFS5 and age to build a LOAD diagnostic model and AUC was 0.723. Results of qRT-PCR experiments with clinical blood samples showed that the three candidate genes were expressed significantly lower in the LOAD and MCI groups when compared to CN. Two mitochondrial-related candidate genes, NDUFA1 and NDUFS5, were identified as diagnostic markers for LOAD and MCI. Combining these two candidate genes with age, a LOAD diagnostic prediction model was successfully constructed.

A risk model based on lncRNA-miRNA-mRNA gene signature for predicting prognosis of patients with bladder cancer.

We aimed to analyze lncRNAs, miRNAs, and mRNA expression profiles of bladder cancer (BC) patients, thereby establishing a gene signature-based risk model for predicting prognosis of patients with BC. We downloaded the expression data of lncRNAs, miRNAs and mRNA from The Cancer Genome Atlas (TCGA) as training cohort including 19 healthy control samples and 401 BC samples. The differentially expressed RNAs (DERs) were screened using limma package, and the competing endogenous RNAs (ceRNA) regulatory network was constructed and visualized by the cytoscape. Candidate DERs were screened to construct the risk score model and nomogram for predicting the overall survival (OS) time and prognosis of BC patients. The prognostic value was verified using a validation cohort in GSE13507. Based on 13 selected. lncRNAs, miRNAs and mRNA screened using L1-penalized algorithm, BC patients were classified into two groups: high-risk group (including 201 patients ) and low risk group (including 200 patients). The high-risk group's OS time ( hazard ratio [HR], 2.160; 95% CI, 1.586 to 2.942; P= 5.678e-07) was poorer than that of low-risk groups' (HR, 1.675; 95% CI, 1.037 to 2.713; P= 3.393 e-02) in the training cohort. The area under curve (AUC) for training and validation datasets were 0.852. Younger patients (age ⩽ 60 years) had an improved OS than the patients with advanced age (age > 60 years) (HR 1.033, 95% CI 1.017 to 1.049; p= 2.544E-05). We built a predictive model based on the TCGA cohort by using nomograms, including clinicopathological factors such as age, recurrence rate, and prognostic score. The risk model based on 13 DERs patterns could well predict the prognosis for patients with BC.

Clinical implications of AGR2 in primary prostate cancer: Results from a large-scale study.

Human anterior gradient-2 (AGR2) has been implicated in carcinogenesis of various solid tumours, but the expression data in prostate cancer are contradictory regarding its prognostic value. The objective of this study is to evaluate the expression of AGR2 in a large prostate cancer cohort and to correlate it with clinicopathological data. AGR2 protein expression was analysed immunohistochemically in 1023 well-characterized prostate cancer samples with a validated antibody. AGR2 expression levels in carcinomas were compared with matched tissue samples of adjacent normal glands. AGR2 expression levels were dichotomized and tested for statistical significance. Increased AGR2 expression was found in 93.5% of prostate cancer cases. AGR2 levels were significantly higher in prostate cancer compared with normal prostate tissue. A gradual loss of AGR2 expression was associated with increasing tumour grade (ISUP), and AGR2 expression is inversely related to patient survival, however, multivariable significance is not achieved. AGR2 is clearly upregulated in the majority of prostate cancer cases, yet a true diagnostic value appears unlikely. In spite of the negative correlation of AGR2 expression with increasing tumour grade, no independent prognostic significance was found in this large-scale study.

Association of Antipsychotic Drugs with Venous Thromboembolism: Data Mining of Food and Drug Administration Adverse Event Reporting System and Mendelian Randomization Analysis.

Past observational studies have reported on the association between antipsychotic drugs and venous thromboembolism (VTE); however, the conclusions remain controversial, and its mechanisms are yet to be fully understood. Thus, in this study, we aim to determine the associations of antipsychotic drugs with VTE, including deep vein thrombosis (DVT) and pulmonary embolism (PE), and their potential mechanisms. We first mined the adverse event signals of VTE, DVT, and PE caused by antipsychotic drugs in Food and Drug Administration Adverse Event Reporting System (FAERS). Next, we used two-sample Mendelian randomization (MR) method to investigate the association of antipsychotic drug target gene expression with VTE, DVT, and PE, using single-nucleotide polymorphisms as genetic instruments. We not only used the expression of all antipsychotic drug target genes as exposure to perform MR analyses but also analyzed the effect of single target gene expression on the outcomes. In the FAERS, 1694 cases of VTE events were reported by 16 drugs. However, using the MR approach, no significant association was determined between the expression of all antipsychotic target genes and VTE, DVT, or PE, either in blood or brain tissue. Although the analysis of single gene expression data showed that the expression of nine genes was associated with VTE events, these targets lacked significant pharmacological action. Adverse event mining results have supported the claim that antipsychotic drugs can increase the risk of VTE. However, we failed to find any genetic evidence for this causal association and potential mechanisms. Thus, vigilance is still needed for antipsychotic drug-related VTE despite the limited supporting evidence.

Genome-wide association study of brain functional and structural networks

Abstract Imaging genetics studies with large samples have identified many genes associated with brain functions and structures, but little is known about genes associated with brain functional and structural network properties. The current genome-wide association study examined graph theory measures of brain structural and functional networks with 497 healthy Chinese participants (17–28 years). Four genes (TGFB3, LGI1, TSPAN18, and FAM155A) were identified to be significantly associated with functional network global efficiency, and two (NLRP6 and ICE2) with structural network global efficiency. Meta-analysis of structural and functional brain network property confirmed the four functional-related genes and revealed two more (RBFOX1 and WWOX). They were reported to be significantly associated with regional brain structural or functional measurements in the UK Biobank project; and showed differential gene expression level between low and high structure–function coupling regions according to Allen Human Brain Atlas gene expression data. Taken together, our results suggest that brain structural and functional networks had shared and unique genetic bases, consistent with the notion of many-to-many structure–function coupling of the brain.

Grouped Spherical Data Modeling Through Hierarchical Nonparametric Bayesian Models and Its Application to fMRI Data Analysis.

Recently, spherical data (i.e., L2 normalized vectors) modeling has become a promising research topic in various real-world applications (such as gene expression data analysis, document categorization, and gesture recognition). In this work, we propose a hierarchical nonparametric Bayesian model based on von Mises-Fisher (VMF) distributions for modeling spherical data that involve multiple groups, where each observation within a group is sampled from a VMF mixture model with an infinite number of components allowing them to be shared across groups. Our model is formulated by employing a hierarchical nonparametric Bayesian framework known as the hierarchical Pitman-Yor (HPY) process mixture model, which possesses a power-law nature over the distribution of the components and is particularly useful for data distributions with heavy tails and skewness. To learn the proposed HPY process mixture model with VMF distributions, we systematically develop a closed-form optimization algorithm based on variational Bayes (VB). The merits of the proposed hierarchical Bayesian nonparametric model for modeling grouped spherical data are demonstrated through experiments on both synthetic data and a real-world application about resting-state functional magnetic resonance imaging (fMRI) data analysis.

Abstract 3490: Unveiling sex differences in lung adenocarcinoma through multi-omics integrative protein signaling networks

Abstract Sex differences in lung adenocarcinoma (LUAD) are evident in incidence rates, prognostic outcomes, and therapy responses, yet the underlying molecular mechanisms driving these disparities remain underexplored. In this study, we conducted a comprehensive proteogenomic analysis encompassing 38 females and 73 males with LUAD from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) dataset. Employing Transcription Inference using Gene Expression and Regulatory data (TIGER), we inferred sex-differentially activated transcription factors (TFs) from The Cancer Genome Atlas (TCGA) LUAD gene expression data and identified sex-differentially activated kinases using CPTAC protein phosphorylation data. We further constructed a comprehensive kinase-TF signaling network by integrating these sex-differentially activated kinases with TFs, identifying all paths shorter than 3 in the protein interaction networks to highlight druggable pathways. Our analyses revealed that many proteins exhibit not only sex-biased abundance but also sex-biased phosphorylation and acetylation. Furthermore, these sex-biased proteins were associated with critical biological pathways including cell proliferation, immune response, and metabolism. Using kinase-TF signaling networks, we found substantial sex bias in the activities of clinically actionable TFs and kinases, including the glucocorticoid receptor (NR3C1), AR, AURKA, CDK6, and MAPK14. Leveraging the PRISM cancer cell line screening database, we identified several small-molecule drugs, such as glucocorticoid receptor agonists and aurora kinase inhibitors, potentially exhibiting sex-specific efficacy as LUAD therapeutics. Our findings showed that the activity of some clinically relevant TFs and kinases differ by sex in LUAD, underscoring the need to consider sex as a biological variable and the utility of multi-omics integrative protein signaling networks in advancing our understanding of cancer biology and the development of sex-aware therapeutics. Citation Format: Chen Chen, Enakshi Saha, Dawn L. DeMeo, John Quackenbush, Camila M. Lopes-Ramos. Unveiling sex differences in lung adenocarcinoma through multi-omics integrative protein signaling networks [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 3490.

Proteomic profiling reveals that ESR1 mutations enhance cyclin-dependent kinase signaling

Three quarters of all breast cancers express the estrogen receptor (ER, ESR1 gene), which promotes tumor growth and constitutes a direct target for endocrine therapies. ESR1 mutations have been implicated in therapy resistance in metastatic breast cancer, in particular to aromatase inhibitors. ESR1 mutations promote constitutive ER activity and affect other signaling pathways, allowing cancer cells to proliferate by employing mechanisms within and without direct regulation by the ER. Although subjected to extensive genetic and transcriptomic analyses, understanding of protein alterations remains poorly investigated. Towards this, we employed an integrated mass spectrometry based proteomic approach to profile the protein and phosphoprotein differences in breast cancer cell lines expressing the frequent Y537N and Y537S ER mutations. Global proteome analysis revealed enrichment of mitotic and immune signaling pathways in ER mutant cells, while phosphoprotein analysis evidenced enriched activity of proliferation associated kinases, in particular CDKs and mTOR. Integration of protein expression and phosphorylation data revealed pathway-dependent discrepancies (motility vs proliferation) that were observed at varying degrees across mutant and wt ER cells. Additionally, protein expression and phosphorylation patterns, while under different regulation, still recapitulated the estrogen-independent phenotype of ER mutant cells. Our study is the first proteome-centric characterization of ESR1 mutant models, out of which we confirm estrogen independence of ER mutants and reveal the enrichment of immune signaling pathways at the proteomic level.

Abstract 6235: Clinical and phenotypic consequences of HLA mediated antigen presentation deficiency in pancreatic adenocarcinoma at bulk and single-cell resolution

Abstract Introduction: Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal neoplasm of the pancreas characterized by a low survival rate and limited treatment options. Despite the success of immunotherapy in various cancer types, its efficacy in PDAC remains low. The underlying reasons for this discrepancy are not fully understood, despite PDAC exhibiting a moderate mutation burden. One possible determinant could be the loss of heterozygosity (LOH) at the human leukocyte antigen (HLA) loci, resulting in compromised antigen presentation. In this work, we aim to identify and validate HLA LOH events in two independent cohorts to gain insights into both their prevalence and clinical and phenotypic impact on PDAC. Methods: We applied LOHHLA, a specialized pipeline to identify HLA LOH events from paired whole genome sequencing to two independent cohorts comprising over 650 patients. We contextualized these results with the paired transcriptome-wide gene expression data and immunohistochemistry stains. In addition, we developed a machine learning classifier to predict HLA LOH from transcriptomic data. We transfer this classifier to single cell RNA sequencing data to identify the impact of (subclonal) HLA LOH on the tumor microenvironment. Results: HLA LOH events occur in approximately 30% of PDAC patients, with about 50% of these deletions being focal deletions. We find evidence that focal but not non-focal LOH is a driver of immune escape. Overall, we find that HLA LOH events are early genetic alterations and observe a significant association between HLA LOH and the Basal PDAC expression subtype. In addition, we find that HLA expression is the most important determinant of lymphocyte infiltration followed by an intact HLA locus. Moreover, our transcriptomic classifier allowed us to accurately identify HLA LOH events from RNA sequencing data, which we successfully validated in an independent cohort. Finally, we applied this classifier to single cell sequencing data and identified phenotypic differences in multiple compartments of the tumor microenvironment between samples with intact HLA and HLA LOH. Conclusions: In conclusion, our study provides the most in depth characterization of the consequences of altered antigen presentation in PDAC to date. We have demonstrated that HLA LOH occurs in a substantial proportion of patients. Overall, these findings contribute to a better understanding of the immune landscape in PDAC and may have implications for the development of immunotherapeutic strategies tailored to this challenging cancer type. Citation Format: Michael J. Geuenich, Barbara Gruenwald, Grainne O'Kane, Chengxin Yu, Tan Tiak Ju, Amy Zhang, Oumaima Hamza, Gun Ho Jang, Faiyaz Notta, Steven Gallinger, Kieran R. Campbell. Clinical and phenotypic consequences of HLA mediated antigen presentation deficiency in pancreatic adenocarcinoma at bulk and single-cell resolution [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 6235.

Abstract 3634: Copy number alterations in chromosome 19 genes associate with reduced immune infiltrate and adverse outcomes in lung cancer

Abstract Deleterious mutations in STK11/LKB1, KEAP1, and SMARCA4 have been shown to be associated with adverse response to immunotherapy in several lung cancer cohorts. However, the biological and clinical implications of deletions of these genes, which all reside on chromosome 19p (chr19p), has not been extensively described. We sought to first characterize the association between immune infiltration and these copy number changes in chr19p. To do this, we analyzed mutation, copy number, and gene expression data from non-small cell lung cancer cohorts in the Cancer Genome Atlas (TCGA). Chr19p is frequently deleted across cancers in the TCGA dataset, with higher rates in lung adenocarcinoma. In both lung adenocarcinoma and squamous cell carcinoma, nearly half of TCGA tumors have deletion of chr19p including STK11, KEAP1, and SMARCA4, and these tumors are enriched for STK11 mutation. Immune infiltrate (estimated by methylation patterns) was not significantly associated with STK11 mutation but was significantly reduced in tumors with chr19 deletion when controlling for tumor type and overall copy number load. In addition, low-level deletion of any of these genes correlated with decreased expression of immune signaling pathways. Based on these findings, we next wanted to characterize the association between chr19p copy number and immunotherapy response. We previously reported that low-level deletion of STK11 correlated with an immunosuppressive, treatment-refractory phenotype in samples collected from non-small cell lung cancer (NSCLC) patients with resectable disease in a phase II study of neoadjuvant atezolizumab + chemotherapy. To confirm these findings in a larger cohort, we examined the AACR GENIE NSCLC biopharma collective (BPC) dataset. A subset of this cohort included patients treated with immunotherapy (atezolizumab, nivolumab, and/or pembrolizumab) and with copy number data available or GISTIC analysis of individual genes (n = 204). Here, deletion of any of these three genes correlated with worse overall survival, and this was statistically significant for STK11 deletion. Overall, our data suggests that a broader set of patients, defined by copy number changes in these genes, may experience this adverse immunobiology. Further study to elucidate the mechanistic implications of this association is warranted. Citation Format: Brian S. Henick, Yohanna Georgis, Benjamin O. Herzberg, Carla P. Concepcion-Crisol, Alison M. Taylor. Copy number alterations in chromosome 19 genes associate with reduced immune infiltrate and adverse outcomes in lung 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 3634.

Abstract 6125: Differential expression of chemokine genes by race and Breast Cancer Consensus Subtypes in hormone receptor positive breast cancer

Abstract Hormone receptor positive (HR+) breast cancer is a heterogeneous subtype and African American women (AAW) are twice as likely to have HR+ tumors with lower levels of HR staining (1-10%) compared to European American women (EAW). These tumors are associated with substantially worse survival compared to strongly HR+ tumors. We further hypothesized that HR+ tumors from AAW would be more likely to have molecular features typically associated with HR-negative breast cancer and poor prognosis. To test this, we used the Breast Cancer Consensus Subtyping (BCCS) method and evaluated expression of chemokine receptors known to be differentially expressed by race, PAM50 subtypes, and survival. We downloaded The Cancer Genome Atlas (TCGA) expression data for 394 female EA and AA HR+/HER2- breast cancer patients using the cBio Cancer Genomics Portal. Breast Cancer Consensus Subtypes (BCCS) were determined using the R program BCCSclassifier. BCCS classification results in five subtypes: ER-negative non-basal-like (BCS1), ER-negative basal-like (BCS2), ER+ highly proliferative with poor prognosis (BCS3), ER+ stromal infiltration with good prognosis (BCS4), and ER+ hormone response genes highly expressed with good prognosis (BCS5). High and low expression of chemokines (ACKR1, ACKR4, CCR3, CCR6, CCRL2, CXCR1, CXCR2, CXCR4, CXCR6, CX3CR1) was dichotomized based on median expression of each gene within our study cohort. Distribution of BCCS by race was assessed with multinomial logistic regression. Associations between chemokine gene expression, BCC subtype and race were evaluated using logistic regression. An alpha value of 0.05 was set to determine statistical significance. From the TCGA cohort, AAW were more likely to have the BCS2 ER-negative basal-like phenotype (OR=7.17, p<0.001) as well as the poor prognosis highly proliferative ER+ BCS3 phenotype (OR=2.83, p=0.036) versus the high hormone expression BCS5 subtype. In multivariable models evaluating both race and BCCS subtype, AAW were 6.17 times more likely to have lower than median expression of CX3CR1 (p<0.001) and CXCR2 (OR=2.23, p=0.0096) compared to EAW. Notably, all chemokine receptors consistently showed significantly lower expression in BCS1-4 subtypes compared to BCS5. For CCRL2, the effect estimates for BCS4 vs. BCS5 significantly differed by race (p=0.039). In AAW, BCS4 was associated with higher than median expression (OR=1.99, p=0.49), although not statistically significant. In EAW, BCS4 was associated with lower than median expression (0.23, p<0.001). Our findings show that AAW with HR+ breast cancer have tumors with more inherently aggressive molecular subtypes associated with poor prognosis compared to EAW. These results also further support the premise that HR+/HER2- tumors are highly heterogeneous and strongly suggest that chemokine gene expression is associated with breast cancer subtype. Citation Format: Abigail M. Fielder, Manohar Ratnam, Gregory Dyson, Kristen Purrington. Differential expression of chemokine genes by race and Breast Cancer Consensus Subtypes in hormone receptor positive breast 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 6125.

Abstract 6775: 3D culture aggregation morphology depends on cell phenotype and gene expression profiles in lung cancer

Abstract Background: Three-dimensional (3D) cell culture allows for the reproduction of tissue structure and function in a state more akin to the in vivo environment compared to two-dimensional (2D) cell culture. It is established that 3D culture more faithfully replicates cell morphology, differentiation, proliferation, gene and protein expression, and responsiveness to stimulation, as well as drug metabolism. Focusing on the observation that lung cancer cell lines exhibit different clustering morphologies in 3D cell culture, we conducted a comparative study on the relationship between these clustering morphologies and the characteristics of the lung cancer cell lines. Methods: We cultured 30 lung cancer cell lines using 3D culture plate, and classified the types of cell clustering morphologies formed. We evaluated the molecular and cellular biology characteristics of all cell lines. Additionally, comprehensive gene expression analysis using RNA sequencing was performed on the 30 3D-cultured cell samples, and the characteristics of gene expression between different clustering types were compared. Results: The 3D culture aggregation morphology of 30 lung cancer cell lines were classified into three types (monolayer sheet, NOS, spheroid). Monolayer sheet-type cells exhibited higher proliferation, migration, and invasion capabilities compared to spheroid-type cells. In the transcriptomic bioinformatics analysis of the acquired gene expression data, GSEA analysis using Hallmark gene sets showed that the enrichment of epithelial-mesenchymal transition (EMT) and cancer metastasis-related genes in the monolayer type cells significantly compared to the spheroid type cells. Conversely, cell adhesion and intercellular binding-related gene sets were highly enriched in the spheroid type cells. Conclusion: 3D culture aggregation morphology represents phenotypic properties and transcriptomic profiles in non-small cell lung cancer cells. Citation Format: Naohiro Hayashi, Shimpei Tsudaka, Ken Suzawa, Tomoaki Higashihara, Yin Min Thu, Atsushi Matsuoka, Daisuke Mizuno, Ryo Yoshichika, Fumiaki Mukohara, Mao Yoshikawa, Kazuhiko Shien, Hiromasa Yamamoto, Shinichi Toyooka. 3D culture aggregation morphology depends on cell phenotype and gene expression profiles in lung 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 6775.

Abstract 5896: Platform approach to develop antibodies specifically recognizing cancer-associated glycoforms

Abstract Introduction: Recent advances in antibody engineering have created a portfolio of highly potent therapeutic approaches like antibody-drug conjugates (ADCs), radioimmunoconjugates (RITs), chimeric antigen receptors (CARs) or bi-specifics. Due to the increased potency, these drugs are more than ever dependent on very clean targets to prevent side effects. We aim to address this major problem through our approach to develop antibodies with improved tumor specificity and reduced on-target/off-tumor binding utilizing the aberrant O-glycosylation on tumor cells. Altered glycosylation of proteins and lipids is one of the most drastic changes in cancer, giving rise to truncated or highly fucosylated and highly sialylated glycans which are almost absent on normal cells. Thus, developing antibodies against protein/carbohydrate combined epitopes (GlycoTargets) comprising these tumor-specific glycans enables highly potent therapies with reduced off-tumor toxicity. Experimental procedures: A workflow and corresponding database was established to evaluate the O-glycosylation of proteins. The first and most basic criterion is the presence of suitable O-glycosylated peptide stretches in the extracellular domain of a protein based on predictions. For proteins fulfilling this criterion cancer-relevant expression is analyzed in a second step. This is done either using publicly available data or experimentally (immunohistochemistry or protein expression data). During the final major step, the theoretical existence of suitable O-glycosylation is confirmed by our own experimental data. This includes analysis of cellularly expressed proteins by anti-glycan antibodies and mass spectrometric studies. Results summary: Case studies will be presented that highlight our workflow to identify suitable GlycoTargets based on publicly available data analysis and experimental confirmation. Additionally, the subsequent development of antibodies against the identified GlycoTargets will be shown. Conclusion: A standardized process was developed for the data collection and prioritization of potential GlycoTargets. The obtained information is subsequently used for targeted discovery of antibodies that bind to protein/carbohydrate combined glycoepitopes (GlycoTargets) offering increased tumor-specificity compared to simple protein targets. Citation Format: Patrik Kehler, Johanna Gellert, Lisa Kalfhues, Sophie Marinoff, Manon Weis, Andreas Franz, Naomi Kast, Stephanie Gurka, Marianne Morche, Evelyn Hartung, Timo Lischke, Sarah Mayer-Hain, Antje Danielczyk. Platform approach to develop antibodies specifically recognizing cancer-associated glycoforms [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 5896.

Comparison of assessment of programmed death-ligand 1 (PD-L1) status in triple-negative breast cancer biopsies and surgical specimens

AimsProgrammed death-ligand 1 (PD-L1) status in triple-negative breast cancer (TNBC) is important for immune checkpoint inhibitor therapies but may vary between different immunohistochemical assays, scorings and the type of specimen...

Bayesian Hyperbolic Multidimensional Scaling

Multidimensional scaling (MDS) is a widely used approach for representing high-dimensional, dependent data. MDS works by assigning each observation a location on a low-dimensional geometric manifold, with distance on the manifold representing similarity. We propose a Bayesian approach to multidimensional scaling when the low-dimensional manifold is hyperbolic. Using hyperbolic space facilitates representing tree-like structures common in many settings (e.g., text or genetic data with hierarchical structure). A Bayesian approach provides regularization that minimizes the impact of measurement error in the observed data and assesses uncertainty. We also propose a case-control likelihood approximation that allows for efficient sampling from the posterior distribution in larger data settings, reducing computational complexity from approximately O ( n 2 ) to O ( n ) . We evaluate the proposed method against state-of-the-art alternatives using simulations, canonical reference datasets, Indian village network data, and human gene expression data. Code to reproduce the result in the article is available at https://github.com/peterliu599/BHMDS. Supplementary materials for this article are available online.