Cancer Cell Line Encyclopedia Research Articles

Abstract A009: Integrative drug screening and multi-omic characterization of patient-derived bladder cancer organoids reveals novel molecular correlates of chemotherapy response

Abstract Introduction: Predicting response to therapy for each patient’s tumor is critical to improving long-term outcomes for muscle-invasive bladder cancer (MIBC). Our objective is to establish ex vivo bladder cancer patient-derived organoid (PDO) models that are representative of patients’ tumors and determine potential efficacy of standard-of-care and curated experimental therapies. Methods: Tumor material was prospectively collected from consented patients with bladder cancer to generate short-term PDO models, which were screened against a panel of clinically relevant drugs in ex vivo 3D culture. Drugs were tested at the maximum plasma concentration (Cmax) in human trials, so as to provide physiologic relevance. When material was sufficient, cells were additionally tested using full dose response. The number of drugs screened was dependent upon amount of tissue available, with up to 34 drugs screened at Cmax and 9 drugs screened in dose response format. Multiomic profiling was utilized to validate the PDO models, establish the molecular characteristics of each tumor, assess gene expression (GEX) patterns between paired primary tissue and PDOs, and identify potential biomarkers of drug response using Pearson correlation coefficients and Kruskal-Wallis tests with Dunn’s post-hoc pairwise comparison testing. Results: 106 tumors were collected from 97 patients, with 65 samples yielding sufficient material for complete multiomic molecular characterization and PDO screening with 6 to 32 drugs/combinations. RNA sequencing and GSEA enriched pathways across organoids were compared to other published datasets, noting high levels of correlation between our organoid models and in particular the Cancer Cell Line Encyclopedia urethral lines. Moreover, analysis of subtype and mutation/copy number data further indicate that our organoid dataset is representative of the full spectrum of disease. RNA sequencing and subtyping of cultured organoids compared to patient tissue indicate that organoid models are representative of patient tissue and do not undergo subtype shifts in our short-term 3D cultures. Our drug screening results highlight the large spectrum of response to chemotherapies that in fact, clinically benefit only a small proportion of patients in the neoadjuvant setting. Utilizing an integrative approach, novel correlations between ex vivo drug responses and genomic alterations, GEX, and protein expression were identified, including a multiomic signature of gemcitabine response. Conclusions: Rapid organoid development, characterization, and drug screening allows for the prediction of therapeutic response in ~10 days following sample collection. Use of this technique on tissue provided during disease work-up may further guide selection of effective therapeutic agents in patients with bladder cancer, beyond the typical unranked lists derived from genomic analyses. This work would help maximize the overall benefits of standard of care therapies and could be used to identify alternative therapeutics for patients who progress on standard therapies. Citation Format: Nathan M. Merrill, Samuel D. Kaffenberger, Liwei Bao, Nathalie Vandecan, Laura E. Goo, Athena Apfel, Xu Cheng, Zhaoping Qin, Chia-Jen Liu, Armand Bankhead, Yin Wang, Varun Kathawate, Lila Tudrick, Habib Serhan, Zackariah Farah, Chad Ellimoottil, Khaled S. Hafez, Lindsey A. Herrel, Jeffrey S. Montgomery, Todd M. Morgan, Simpa S. Salami, Alon Z. Weizer, Peter J. Ulintz, Mark L. Day, Matthew B. Soellner, Phillip L. Palmbos, Sofia D. Merajver, Aaron M. Udager. Integrative drug screening and multi-omic characterization of patient-derived bladder cancer organoids reveals novel molecular correlates of chemotherapy response [abstract]. In: Proceedings of the AACR Special Conference on Bladder Cancer: Transforming the Field; 2024 May 17-20; Charlotte, NC. Philadelphia (PA): AACR; Clin Cancer Res 2024;30(10_Suppl):Abstract nr A009.

GJB3: a comprehensive biomarker in pan-cancer prognosis and immunotherapy prediction.

A wide range of connexins are situated between normal-normal cells, cancer-cancer cells, and cancer-normal cells. Abnormalities in connexin expression are typically accompanied by cancer development; however, no systematic studies have examined the role of Gap Junction Protein Beta 3 (GJB3) in the context of tumor progression and immunity, especially when considering a broad range of cancer types. In this study, data on GJB3 expression were gathered from Genotype-Tissue Expression, Cancer Cell Line Encyclopedia, and The Cancer Genome Atlas databases. Then, we analyzed the relationship between GJB3 expression and tumor characteristics. In vitro experiments using colony formation, EdU, CCK8, transwell migration assays, immunohistochemistry and western blot were performed to investigate the function of GJB3 in tumor progression of various cell lines. A drug sensitivity analysis of GJB3 was performed using the Genomics of Drug Sensitivity in Cancer database. Our findings demonstrate that GJB3 is widely expressed in various cancers and correlates significantly with disease stages, patient survival, immunotherapy response, and pharmaceutical guidance. Additionally, GJB3 plays a role in different cancer pathways, as well as in different immune and molecular subtypes of cancer. Co-expression of GJB3 with immune checkpoint genes was observed. Further experiments showed that knockdown of GJB3 inhibited the PI3K/AKT pathway and resulted in reduced proliferation, migration, and viability of different cancer cells. Overall, GJB3 shows potential as a molecular biomarker and therapeutic target for various cancers, particularly lung adenocarcinomas, mesothelioma, pancreatic adenocarcinoma. Thus, GJB3 may represent a new therapeutic target for a wide range of cancers.

HANSynergy: Heterogeneous Graph Attention Network for Drug Synergy Prediction.

Drug synergy therapy is a promising strategy for cancer treatment. However, the extensive variety of available drugs and the time-intensive process of determining effective drug combinations through clinical trials pose significant challenges. It requires a reliable method for the rapid and precise selection of drug synergies. In response, various computational strategies have been developed for predicting drug synergies, yet the exploitation of heterogeneous biological network features remains underexplored. In this study, we construct a heterogeneous graph that encompasses diverse biological entities and interactions, utilizing rich data sets from sources, such as DrugCombDB, PubChem, UniProt, and cancer cell line encyclopedia (CCLE). We initialize node feature representations and introduce a novel virtual node to enhance drug representation. Our proposed method, the heterogeneous graph attention network for drug-drug synergy prediction (HANSynergy), has been experimentally validated to demonstrate that the heterogeneous graph attention network can extract key node features, efficiently harness the diversity of information, and further enhance network functionality through the incorporation of a multihead attention mechanism. In the comparative experiment, the highest accuracy (Acc) and area under the curve (AUC) are 0.877 and 0.947, respectively, in DrugCombDB_early data set, demonstrating the superiority of HANSynergy over the competing methods. Moreover, protein-protein interactions are important in understanding the mechanism of action of drugs. The heterogeneous attention mechanism facilitates protein-protein interaction analysis. By analyzing the changes of attention weight before and after heterogeneous network training, we investigated proteins that may be associated with drug combinations. Additionally, case studies align our findings with existing research, underscoring the potential of HANSynergy in drug synergy prediction. This advancement not only contributes to the burgeoning field of drug synergy prediction but also holds the potential to provide valuable insights and uncover new drug synergies for combating cancer.

Integration of Single-Cell RNA Sequencing and Bulk RNA Sequencing to Identify an Immunogenic Cell Death-Related 5-Gene Prognostic Signature in Hepatocellular Carcinoma.

Immunogenic cell death (ICD) can enhance the potency of immunotherapy in cancer treatment. Nevertheless, it is ambiguous how ICD-related genes (ICDRGs) contribute to hepatocellular carcinoma (HCC). Single-cell RNA sequencing (scRNA-seq) data were used to distinguish malignant cells from normal cells in the HCC tumor microenvironment(TME). Bulk RNA sequencing data was employed to acquire the landscape of the 33 ICDRGs. Unsupervised clustering identified two ICD molecular subtypes. The cellular infiltration characteristics and biological behavior in different subtypes were analyzed by ssGSEA. Subsequently, differentially expressed genes (DEGs) between the two subtypes were determined, based on which patients were classified into three gene clusters. Then, the prognostic model was constructed by Lasso-Cox analysis. Finally, we investigated the expression of risk genes in cancer cell line encyclopedia (CCLE) and validated the function of NKX3-2 in vitro experiments. ICD scores and ICDRGs expression in malignant cells were significantly lower than in normal cells by scRNA-seq analysis. ICD-high subtype was characterized by ICD-related gene overexpression and high levels of immune infiltration abundance and immune checkpoints; Three DEGs-related gene clusters were likewise strongly linked to stromal and immunological activation. In the ICD-related prognostic model consisting of NKX3-2, CHODL, MMP1, NR0B1, and CTSV, the low-risk group patients had a better endpoint and displayed increased susceptibility to immunotherapy and chemotherapeutic drugs like 5-Fluorouracil, afatinib, bortezomib, cediratinib, lapatinib, dasatinib, gefitinib and crizotinib. Moreover, NKX3-2 amplification in HCC samples has been verified by experiments, and its disruption suppressed the proliferation and invasion of tumor cells. Our study highlighted the potential of the ICDRGs risk score as a prognostic indicator to aid in the accurate diagnosis and immunotherapy sensitivity of HCC.

Glioma hexokinase 3 positively correlates with malignancy and macrophage infiltration.

Glioma is the main subtype of primary central nervous system (CNS) tumor with high malignancy and poor prognosis under current therapeutic approaches. Glycolysis and suppressive tumor microenvironment (TME) are key markers of glioma with great importance for aggressive features of glioma and inferior clinical outcomes. Hexokinase 3 (HK3) is an important rate-limiting enzyme in glycolysis, but its function in glioma remains unknown. This study comprehensively assessed the expression distribution and immunological effect of HK3 via pan-cancer analysis based on datasets from Genotype Tissue Expression (GTEx), Cancer Cell Line Encyclopedia (CCLE), and The Cancer Genome Atlas (TCGA). Furthermore, it explored the malignant phenotype and genomic landscape between low-HK3 and high-HK3 expression groups in gliomas from Chinese Glioma Genome Atlas (CGGA) and TCGA. Moreover, data from the TIMER website predicted the relationship between macrophage infiltration and HK3 expression. Also, single-cell sequencing data were used to validate the relationship. For pan-cancer patients, HK3 was expressed in various cancers. The results showed that HK3 was highly expressed in gliomas and positively correlated with tumor-infiltrating immune cells (TIICs), immune checkpoints, immunomodulators, and chemokines. Meanwhile, HK3 expression was highest in normal immune cells and tissues. In gliomas, the expression of HK3 was found to be closely correlated with the malignant clinical characteristics and the infiltration of macrophages. Also, HK3 was proven to be positively associated with macrophage through single-cell sequencing data and immunohistochemistry techniques. Finally, it is predicted that samples with high HK3 expression are often malignant entities and also significant genomic aberrations of driver oncogenes. This is the first comprehensive research to figure out the relationship between HK3 and TME characteristics in gliomas. HK3 is positively associated with macrophage infiltration and can induce the immunosuppressive TME and malignant phenotype of gliomas.

Identification of genes related to ribosomal proteins in colorectal cancer: exploring their potential as biomarkers, prognostic indicators, and therapeutic targets.

Colorectal cancer (CRC) ranks as the third most commonly diagnosed cancer in both females and males, underscoring the need for the identification of effective biomarkers. We assessed the expression levels of ribosomal proteins (RPs) at both mRNA and protein levels. Subsequently, leveraging the STRING database, we constructed a protein-protein interaction network and identified hub genes. The co-expression network of differentially expressed genes associated with CRC and their target hub RPs was constructed using the weighted gene co-expression network analysis algorithm. Gene ontology and molecular signatures database were conducted to gain insights into the biological roles of genes associated with the identified module. To confirm the results, the expression level of the candidate genes in the CRC samples compared to the adjacent healthy was evaluated by the RT-qPCR method. Our findings indicated that the genes related to RPs were predominantly enriched in biological processes associated with Myc Targets, Oxidative Phosphorylation, and cell proliferation. Also, results demonstrated that elevated levels of GRWD1, MCM5, IMP4, and RABEPK that related to RPs were associated with poor prognostic outcomes for CRC patients. Notably, IMP4 and RABEPK exhibited higher diagnostic value. Moreover, the expression of IMP4 and RABEPK showed a significant association with drug resistance using cancer cell line encyclopedia and genomics of drug sensitivity in cancer databases. Also, the results showed that the expression level of IMP4 and RABEPK in cancerous samples was significantly higher compared to the adjacent healthy ones. The general results of this study have shown that many genes related to RPs are increased in cancer and could be associated with the death rate of patients. We also highlighted the therapeutic and prognostic potentials of RPs genes in CRC.

A hybrid stacking classifier with feature selection for handling imbalanced data

Nowadays, cancer has become more alarming. This paper discusses the most significant Ovarian Cancer, Epithelial Ovarian Cancer (EOC), due to the low survival rate. The proposed algorithm for this work is a ‘Multi classifier ShapRFECV based EOC’ (MSRFECV-EOC) subtype analysis technique that utilized the EOC data from the National Centre for Biotechnology Information and Cancer Cell Line Encyclopedia websites for early identification of EOC using Machine Learning Techniques. This approach increases the data size, balances different classes of the data, and cuts down the enormous number of features unrelated to the disease of interest to prevent overfitting. To incorporate these functionalities, in the data preprocessing stage, OC-related gene names were taken from the Cancermine database and other OC-related works. Moreover, OC datasets were merged based on OC genes, and missing values of EOC subtypes were identified and imputed using Iterative Logistic Imputation. Synthetic Minority Oversampling Technique with an Edited Nearest Neighbors approach is applied to the imputed dataset. Next, in the Feature Selection phase, the most significant features for subtypes of EOC were identified by applying the Shapley Additive Explanations based on the Recursive Feature Elimination Cross-Validation (ShapRFECV) algorithm, preserving predefined features while selecting new EOC features. Eventually, an accuracy of 97% was achieved with Optuna-optimized Random Forest, which outperformed the existing models. SHAP plotted the most prominent features behind the classification. The Pickle tool saves much training time by preserving hidden parameter values of the model. In the final phase, by using the Stratified K Fold Stacking Classifier, the accuracy was improved to 98.9%.

The prognostic value of MEK pathway-associated estrogen receptor signaling activity for female cancers.

Other than for breast cancer, endocrine therapy has not been highly effective for gynecologic cancers. Endocrine therapy resistance in estrogen receptor positive gynecologic cancers is still poorly understood. In this retrospective study, we examined the estrogen receptor (ER) signaling pathway activities of breast, ovarian, endometrial, and cervical cancers to identify those that may predict endocrine therapy responsiveness. Clinical and genomic data of women with breast and gynecological cancers were downloaded from cBioPortal for Cancer Genomics. Estrogen receptor alpha (ESR1) expression level and sample-level pathway enrichment scores (EERES) were calculated to classify patients into four groups (low/high ESR1 and low/high EERES). Correlation between ESR1/EERES score and survival was further validated with RNAseq data from low-grade serous ovarian cancer. Pathway analyses were performed among different ESR1/EERES groups to identify genes that correlate with endocrine resistance, which are validated using Cancer Cell Line Encyclopedia gene expression and Genomics of Drug Sensitivity in Cancer data. We identified a novel combined prognostic value of ESR1 expression and the corresponding estrogen response signaling (EERES score) for breast cancer. The combined prognostic value (ESR1/EERES) may be applicable to other gynecologic cancers. More importantly, we discovered that ER signaling can cross-regulate MEK pathway activation. We identified downstream genes in the MEK pathway (EPHA2, INAVA, MALL, MPZL2, PCDH1, and TNFRSF21) that are potential endocrine therapy response biomarkers. This study demonstrated that targeting both the ER and the ER signaling activity related MEK pathway may aid the development of endocrine therapy strategies for personalized medicine.

Abstract LB247: TCPAplus: An LLM-empowered chatbot for analyzing a large protein expression atlas of human cancers

Abstract We present a comprehensive functional proteomics resource comprising nearly 9,000 samples from The Cancer Genome Atlas and Cancer Cell Line Encyclopedia, using ~500 high-quality antibodies through reverse phase protein arrays. To enhance its utility, we have developed an analytical chatbot harnessing advanced large language models. This intuitive platform empowers researchers to effortlessly query protein-centered cancer omics data, conduct related analyses, and visualize results using natural language, significantly easing the translation from intricate proteogenomic data to valuable biological insights. Citation Format: Wei Liu, Jun Liu, Yitao Tang, Chaozhong Liu, Meiyi Song, Zhenlin Ju, Shweth V. Kumar, Yiling Lu, Rehan Akbani, Gordon Mills, Han Liang. TCPAplus: An LLM-empowered chatbot for analyzing a large protein expression atlas of human cancers [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 LB247.

Abstract NG04: Targeting DHX9 to trigger viral mimicry and immunotherapy responsiveness in small cell lung cancer

Abstract NG04: Targeting DHX9 to trigger viral mimicry and immunotherapy responsiveness in small cell lung cancer

DBDNMF: A Dual Branch Deep Neural Matrix Factorization method for drug response prediction.

Anti-cancer response of cell lines to drugs is in urgent need for individualized precision medical decision-making in the era of precision medicine. Measurements with wet-experiments is time-consuming and expensive and it is almost impossible for wide ranges of application. The design of computational models that can precisely predict the responses between drugs and cell lines could provide a credible reference for further research. Existing methods of response prediction based on matrix factorization or neural networks have revealed that both linear or nonlinear latent characteristics are applicable and effective for the precise prediction of drug responses. However, the majority of them consider only linear or nonlinear relationships for drug response prediction. Herein, we propose a Dual Branch Deep Neural Matrix Factorization (DBDNMF) method to address the above-mentioned issues. DBDNMF learns the latent representation of drugs and cell lines through flexible inputs and reconstructs the partially observed matrix through a series of hidden neural network layers. Experimental results on the datasets of Cancer Cell Line Encyclopedia (CCLE) and Genomics of Drug Sensitivity in Cancer (GDSC) show that the accuracy of drug prediction exceeds state-of-the-art drug response prediction algorithms, demonstrating its reliability and stability. The hierarchical clustering results show that drugs with similar response levels tend to target similar signaling pathway, and cell lines coming from the same tissue subtype tend to share the same pattern of response, which are consistent with previously published studies.

Abstract 6210: The cancer omics and drug experimental response dataset (CODERData): A harmonized benchmark dataset for machine learning models of drug response prediction

Abstract Background: Determining a patient’s response to a specific therapy is a vital step in developing personalized cancer treatment. Personalized treatment relies on two key technological advancements: algorithms tailored to each patient’s molecular profiles, and comprehensive datasets detailing the effects of patient diversity on drug response. Despite recent advancements in machine learning-based models, each new algorithm requires curation of numerous datasets, which poses a significant burden to researchers who need to evaluate and compare various algorithms. Currently, datasets include extensive genomics, transcriptomics, proteomics and metabolomics measurements along with drug response data (i.e., cell viability) for each cell line, collected through the Cancer Cell Line Encyclopedia (CCLE) and other initiatives. Additional omics data is available for tumor organoids and patient samples generated by consortia such as Human Cancer Model Initiative (HCMI) and the Clinical Proteomic Tumor Analysis Consortium (CPTAC). However, to date there has been little public effort to annotate and harmonize these data across repositories. Methods: To address this need, we developed the Python package coderdata that collates the most up-to-date cancer omics and drug response data. We curated four public cancer datasets, establishing reproducible pipelines that standardize data retrieval, formatting, and integration so the datasets can be built on a cluster if needed. The datasets undergo bi-monthly synchronization with FigShare, ensuring the data is up to date and available. Scientists can use coderdata to retrieve and format data on the command line interface as well as within Python scripts. Results: Coderdata is a straightforward Python package that will enable the development and benchmarking of diverse machine learning applications in cancer research. At its core, it represents reproducible data harmonization that enables heterogeneous datasets to be analyzed in bulk. This Python package downloads and standardizes data from multiple consortia-related resources including HCMI, CPTAC, CCLE, and BeatAML, together representing 4931 multi-omics samples from across 250+ cancer types. These datasets include cancer data metrics such as copy number, mutation, transcriptomics, proteomics, miRNA, methylation, and drug response data. Conclusions: The applications of machine learning in cancer biology are hampered by the distributed nature of existing datasets. As such, the collection and standardization of data by coderdata substantially reduces the time investment required for data curation in cancer research. Direct access to benchmark multi-omics and drug response datasets enables scientists to focus on algorithm development for tumor drug response prediction, potentially accelerating the discovery of therapeutic strategies. Citation Format: Jeremy Jacobson, Sydney Schwartz, M. Ryan Weil, Neeraj Kumar, Sara Gosline. The cancer omics and drug experimental response dataset (CODERData): A harmonized benchmark dataset for machine learning models of drug response prediction [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 6210.

Abstract 5829: Elucidating the significance of amphiregulin and its implications for EGFR inhibition strategies in LKB1-mutant NSCLC

Abstract LKB1-mutant non-small cell lung cancer (NSCLC) is aggressive and treatment-refractory, underscoring the need for new therapeutic strategies. Our comparative study involving mouse lung tumors driven by oncogenic KrasG12D and LKB1 loss as well as tumors driven solely by KrasG12D, revealed that LKB1 loss upregulated amphiregulin (Areg), a pro-growth EGFR ligand. Therefore, this study investigates how the inactivation of LKB1 enhances AREG expression, the significance of elevated AREG in sustaining lung cancer growth, and the responses of AREG-high lung cancer to EGFR inhibitors. We first evaluated the potential clinical significance of AREG using The Cancer Genome Atlas (TCGA) lung adenocarcinoma dataset and the Cancer Cell Line Encyclopedia (CCLE) database. We observed that LKB1-null tumors with high AREG levels were associated with shorter patient survival and exhibited transcriptional signatures similar to EGFR-mutant tumors. High AREG expression also correlated positively in NSCLC cell lines with LKB1 mutations and Western blot analysis confirmed that elevated AREG levels in lung cancer cells were associated with constitutive EGFR signaling. We next explored the impact of LKB1 loss on AREG expression in lung cancer. Our prior RNA-seq and ChIP-seq investigations demonstrated CRTC2 and CREB enrichment on the AREG promoter in LKB1-null lung cancer cells. Upon LKB1 reintroduction, the enrichment diminished, indicating that CRTC-CREB transcriptionally induces AREG in LKB1-null lung cancer. To corroborate, reintroducing LKB1 or expressing a dominant-negative construct (dnCRTC), disrupting CREB/CRTC interaction, decreased AREG levels, as confirmed by RT-qPCR in LKB1-null lung cancer cells. This supports that AREG is a direct target of CRTC-CREB activation induced by LKB1 loss. To assess the impact of elevated AREG in LKB1-mutant lung cancer growth, we used RNA interference and antibodies to inhibit AREG. Both interventions decreased cell growth and colony formation, underscoring AREG is critical for LKB1-mutant lung cancer growth. Moreover, we investigated the potential correlation between AREG and the response to the EGFR tyrosine-kinase inhibitor Erlotinib in NSCLC cell lines, utilizing AREG expression data from CCLE and Erlotinib response data from the Genomics of Drug Sensitivity in Cancer (CDSC) databases. Stratifying NSCLC cell lines into sensitive, intermediate, and resistant categories revealed a potential positive association between AREG expression and Erlotinib sensitivity. Notably, LKB1-null lung cell lines with high AREG expression demonstrated heightened sensitivity compared to those with low AREG expression. Our study revealed that LKB1 loss increases AREG, triggering aberrant EGFR signaling critical for lung cancer growth, and AREG expression might impact the Erlotinib response in LKB1-inactive cells. Citation Format: Xzaviar Kaymar Solone, Xin Zhou, Mu Yu, Jennifer W. Li, Finn Rimay, Frederic Kaye, Lizi Wu. Elucidating the significance of amphiregulin and its implications for EGFR inhibition strategies in LKB1-mutant NSCLC [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 5829.

Abstract 1753: SHOC2 is a genetic dependency in NRAS-mutant melanoma

Abstract Melanoma is the most lethal form of skin cancer, with over 300,000 new cases worldwide every year. Specifically, melanoma with NRAS mutations are of particular clinical concern due to their association with a poor prognosis and lack of specific treatment options. Therefore, there is a pressing need for novel approaches to address the treatment of NRAS-mutant melanoma. A current approach to identify novel drug targets is based on the genetic concept of induced essentiality, where functional interactions that occur in response to oncogene addiction create a dependency on another gene. To identify genetic dependencies in NRAS-mutant melanoma, whole-genome CRISPR-Cas9 knockout screens were conducted in 6 NRAS-mutant and 7 NRAS-wild-type (wt) early-passage New Zealand melanoma (NZM) cell lines that were established and maintained at 5% oxygen to mimic physiological oxygen conditions. The NZM cell lines were stably transduced with the whole-genome Brunello lentiviral single guide (sg) RNA library and screened for up to 35 days. BAGEL (Bayesian Analysis of Gene Essentiality) analyses of the NZM whole-genome knockout screens, alongside CRISPR-Cas9 screening data using the Avana sgRNA library from an additional 28 melanoma cell lines, available on the Cancer Cell Line Encyclopaedia (CCLE) database, revealed prospective gene candidates that exhibit greater detrimental effects on the fitness of NRAS-mutant cell lines compared to the NRAS-wt lines. These genes are being further validated as essential genes for NRAS-mutant melanoma cells through custom sgRNA library knockout screens and in vitro individual gene knockout studies. In particular, we demonstrate that knockout of SHOC2, a gene that encodes a scaffold protein essential for activation of the MAPK signalling pathway, results in the prevention of ERK phosphorylation and a more substantial reduction in cell growth in NRAS-mutant NZM cell lines when compared to NRAS-wt lines. These results support previous studies that have identified SHOC2 as a potential therapeutic target for RAS-driven cancers and suggest that targeting of SHOC2 may have utility in NRAS-mutant melanoma, where greater treatment options are urgently needed. Citation Format: Andrea Y. Gu, Tet-Woo Lee, Aziza Khan, Francis W. Hunter, Dean C. Singleton, Stephen M. F. Jamieson. SHOC2 is a genetic dependency in NRAS-mutant melanoma [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 1753.

Abstract 862: Multi-omic analysis of allelic expression patterns of imprinted and non-imprinted genes in cancer and their association with drug response

Abstract Malignant cells undergo a broad array of epigenetic changes, including changes in imprinting patterns. In normal cells, imprinted genes have parent of origin-specific monoallelic expression, which is controlled via epigenetic mechanisms. Imprinted gene loci are often complex and frequently include multiple isoforms, which have different individual patterns of parent-of-origin monoallelic or biallelic expression in the normal tissues. Many imprinted genes promote cell proliferation and body growth, and their expression is important during embryonic and postnatal development. Dysregulation of imprinting in tumors has been suggested to contribute to growth and proliferation of cancer cells. Previously we reported that chemotherapy responses of cancer cell lines from a variety of tumors and of patient-derived acute myeloid leukemia samples were associated with copy number, expression, or DNA methylation of selected imprinted genes. In the current project, we examined allele-specific patterns of expression (monoallelic vs. biallelic) of imprinted genes and isoforms in tumors and studied their potential effect on response of cancer cells to drug treatment. We analyzed expression patterns at the whole gene, isoform, and exon levels of 59,283 autosomal protein-coding genes and ncRNA genes, with a special focus on 94 imprinted genes and their isoforms. We developed a computational pipeline to evaluate monoallelic and biallelic expression patterns of each gene in cancer cell lines and in tumor samples based on available matched RNA-seq expression data, whole exome sequencing information, and copy number data. Our initial analysis included 108 cell lines from the Cancer Cell Line Encyclopedia (CCLE), which had been derived from 9 pediatric and adult tumor histologies. We observed variation of allelic expression patterns of imprinted genes and their isoforms in tumor cell lines, including differences across and within cancer histologies. We also observed predominantly monoallelic expression of multiple additional genes. Some of them (PRIM2, BCLAF1, MAP2K3, and SEC22B) had been previously reported as potentially imprinted. Monoallelic expression of other genes identified in our analysis (e.g., GRK2, CBX4, and additional genes) was likely caused by molecular mechanisms other than imprinting. Allelic expression patterns and overall expression levels of isoforms of multiple imprinted and non-imprinted monoallelically expressed genes (e.g., CPA4, MAP2K3, and other genes) were associated with in vitro drug response. We also report our ongoing validation of allelic expression patterns of imprinted and non-imprinted genes, isoforms, and exons in primary bladder and breast tumors and in matched solid normal tissue samples using patient data from the Cancer Genome Atlas (TCGA). These results provide new knowledge about epigenetic dysregulation in cancer. Citation Format: Julia Krushkal, Travis L. Jensen, George Wright, Yingdong Zhao. Multi-omic analysis of allelic expression patterns of imprinted and non-imprinted genes in cancer and their association with drug response [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 862.

Abstract 3042: Unraveling the mechanism of YAP toxicity in neuroendocrine solid tumors: Insights from Merkel cell carcinoma

Abstract The Hippo pathway, a pivotal regulator orchestrating organ size and tissue growth, has emerged as a central player in cancer biology. The downstream transcription factor YAP within this pathway has been implicated in diverse oncogenic processes. YAP activation contributes to drug resistance, metastasis, and poor clinical outcomes in many solid tumors. Conversely, YAP exhibits growth-suppressive effects in neuroendocrine (NE) solid cancers, a paradox that remains poorly understood. Analysis of RNAseq data from the Cancer Cell Line Encyclopedia (CCLE) revealed a distinct classification of tumors into YAP1-expressing (YAPon) and non-expressing (YAPoff) subgroups, with hematopoietic malignancies and NE solid tumors featured prominently in the latter. Merkel cell carcinoma (MCC) is a rare, high-grade, NE carcinoma of the skin with two distinct etiologies. Virus-positive MCC contains integrated Merkel cell polyomavirus that expresses the viral T antigens and virus-negative MCC caused by sunlight exposure and extensive UV damage. Bulk RNAseq performed with more than 60 MCC tumors revealed that YAP expression displayed an inverse correlation with an NE gene expression signature. To gain insight into why YAP expression is so low in MCC, we generated MCC cell lines with inducible expression of YAP. Expression of YAP led to growth arrest and cell death. We performed a genome-wide CRISPR-Cas9 screen to identify genes that when lost enable MCC cells to tolerate YAP expression. Functional enrichment analysis of the positively selected genes uncovered a variety of signaling pathways. Intriguingly, the CRISPR screen revealed a positive enrichment of non-canonical BAF complex members, including BRD9 and SMARCD1. To validate this finding, we treated MKL-1 cells with a BRD9 inhibitor, observing a substantial reduction in YAP toxicity. These findings open avenues for the development of targeted therapeutic strategies for high grade NE cancers, addressing a critical gap in our understanding of YAP biology in the context of MCC and beyond. Citation Format: Furkan Bahar, Thomas C. Frost, Varsha Ananthapadmanabhan, Julia L. Schnabel, Manisha Thakuria, James A. DeCaprio. Unraveling the mechanism of YAP toxicity in neuroendocrine solid tumors: Insights from Merkel cell carcinoma [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 3042.

Abstract 3502: Multi-omics approach to cholangiocarcinoma-related biology and development of precision medicine

Abstract Introduction: Cholangiocarcinoma is a diverse tumor entity characterized by distinct tumor biology and clinical phenotypes, which are influenced by the unique tumor microenvironment (TME) and anatomical location—whether intrahepatic, peri-hilar, or distal extrahepatic. Consequently, there is a clear need for a precision strategy that relies on cancer-specific, multi-omic features, moving away from traditional molecular subtypes derived from bulk RNA sequencing. Methods: We conducted thorough multi-omic analyses utilizing data from the Cancer Dependency Map (DepMap) project. This encompassed cancer-specific molecular characterization through various omics data, genome-wide loss-of-function screening employing the CRISPR-Cas9 system, and drug sensitivity assessments via compound screening. These efforts aimed to reveal cancer-specific molecular subtypes with clinical relevance. The identified cancer-specific molecular signatures were subsequently validated in independent translational cohorts. Results: Through integrative profiling of the transcriptome in bile duct cancer cell lines (n=39) obtained from the Cancer Cell Line Encyclopedia (CCLE), we unveiled cancer-specific molecular subtypes exhibiting distinct tumor biology across all omics layers, coupled with clinical relevance and prognostic significance. The major molecular features of each subtype demonstrated reproducibility in validation cohorts. Additionally, subtype-specific molecular biomarkers, encompassing mutational signatures and metabolites, were identified. Finally, we analyzed target drugs with subtype-specific genetic dependencies to formulate a precision strategy tailored to the distinct molecular characterizations of each subtype. Conclusions: A thorough analysis utilizing a multi-omics dataset uncovered precision strategies founded on cancer-specific molecular subtypes of cholangiocarcinoma, both in terms of tumor classification and discriminative therapeutic opportunities. It is imperative to conduct prospective translational studies in conjunction with clinical trials, all centered around the cancer-specific molecular subtypes, to establish a precision strategy for effectively managing cholangiocarcinoma. Citation Format: Woo Young Kwon, Incheon Kang, Sung Hwan Lee. Multi-omics approach to cholangiocarcinoma-related biology and development of precision medicine [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 3502.

Abstract 3527: Application of a deep learning based drug sensitivity prediction model on a novel anticancer drug

Abstract Discovery of a candidate drug molecule through the preclinical stages is an iterative multiparametric process until a candidate is selected for investigational new drug submission enabling studies. This iterative approach takes up significant time and effort to design, synthesize, and test new compounds. A specific testing step is evaluation of half maximal inhibitory concentration (IC50) of compounds on cell line (CL) models. To augment CL testing, computational approaches predicting the IC50 of a drug can be used to guide choice of optimization as the molecule goes through hit to lead optimization onto candidate selection. DeepDSC [1] is a deep learning (DL) model developed to predict the IC50 in a specific CL of a given drug by integrating public data from large-scale drug screens and high-throughput RNAseq profiling. The Cancer Cell Line Encyclopedia (CCLE) [2] and Genomics of Drug Sensitivity (GDSC) [3] projects compiled pharmacological profiles of 23 and 139 drugs across 491 and 655 cancer CLs, respectively. A 10-fold cross validation (CV) approach was employed to train DeepDSC on the datasets independently and reported the average root mean squared error (RMSE) and the coefficient of determination (R2) on the held-out sets across the folds. However, the model’s generalizability to novel compounds not in the dataset, specifically compounds that are structurally dissimilar, is yet to be tested. In this regard, the DeepDSC model was retrained, with modifications, on a subset of the GDSC dataset - GDSC2 - to access the generalizability of DL models on BRG399, a proprietary microtubule targeting agent (MTA) developed by BPGbio Inc. GDSC2 was used since data generated for BRG399 and GDSC2 share the same potency assay (CellTiter-Glo [6]), The GDSC2 set was split into training and held-out sets using a 80-20 random split. Further, we chose to optimize the learning rate for both the autoencoder and feed forward networks (FFN), and the L2 regularization factor for the FFN only. Hyperparameters were optimized using 5-fold CV on the training set. Finally, we evaluated the predictive performance of the model on BRG399, which was not present in the training set. We encoded each compound in GDSC2 (n=234) using Circular Fingerprints [4] and observed an average Jaccard (or Tanimoto) distance (JD) [5] of 0.892 across 27261 drug pairs. BRG399 showed an average JD of 0.873 across 234 drug pairs, with the most similar compound JD of 0.785. Our retrained DeepDSC model had RMSE 1.15 and R2 0.82 on the training set and RMSE 1.19 and R2 0.86 on the held-out set. However, on BRG399 tested CLs (n=101) model had RMSE 1.84 and R2 < 0, which indicated that predictions were on average worse compared to a mean-predictor model for BRG399. These results suggest that DL models that can generalize predictions to a held-out set for compounds that were present in the training set may not accurately predict IC50 of novel compounds not in the training set. Citation Format: Nischal M. Chand, Vivek Vishnudas, Michael A. Kiebish, Anjan Thakurta, Niven R. Narain, Stephane Gesta, Gregory M. Miller. Application of a deep learning based drug sensitivity prediction model on a novel anticancer drug [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 3527.

Abstract 7348: CytoCellDB: A gold standard database for classification and analysis of extrachromosomal DNA in cancer

Abstract Extrachromosomal DNA (ecDNA), or double minute chromosomes, are established cytogenetic markers for malignancy and genome instability. More recently, the cancer community has gained a heightened awareness of the roles of ecDNA in cancer proliferation, drug resistance and epigenetic remodeling. A current hindrance to understanding the biological roles of ecDNA is the lack of available cell line model systems with experimental cytogenetic data that confirm ecDNA status. Although several recent landmark studies have identified numerous cell lines and tumor models with ecDNA, the current sample size limits our ability to detect ecDNA-driven molecular differences due to limitations in power. Increasing the number of model systems that are known to express ecDNA would provide new avenues for understanding the fundamental underpinnings of ecDNA biology and would unlock a wealth of potential targeting strategies for ecDNA-driven cancers. To bridge this gap, we created CytoCellDB, a resource that provides karyotype annotations, including ecDNA status, and leverages publicly available global cell line data from the Cancer Dependency Map (DepMap) and the Cancer Cell Line Encyclopedia (CCLE). We expanded the total number of cancer cell lines with ecDNA annotations to 632, or 34% of the cell lines in CCLE/DepMap. Of these, 154 cell lines express ecDNA, which is a 240% increase from the current sample size. We demonstrate that a strength of our framework is the ability to interrogate ecDNA, and the compendium of other chromosomal aberrations, in the context of cancer-specific vulnerabilities, drug sensitivities, and other multi-omic features (genomics, transcriptomics, methylation, proteomics). We anticipate that this resource will lower the barrier to systematic, population-scale discoveries related to ecDNA and will provide insights into strategies and best practices for determining novel therapeutics that overcome ecDNA-mediated drug resistance. Citation Format: Liz Brunk. CytoCellDB: A gold standard database for classification and analysis of extrachromosomal DNA in 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 7348.

Abstract 4351: Differential dependency mapping of chimeric RNAs across cancer reveals a new landscape of functional fusion transcripts

Abstract Chimeric RNAs are RNA transcripts containing sequences originating from multiple distinct genetic loci and can form through a variety of mechanisms such as DNA rearrangement, cis-splicing of adjacent genes, or RNA trans-splicing. Chimeras have long been known to be a hallmark of cancer and due to their unique properties are promising targets for precision medicine. Appropriately, fusion proteins transcribed from chimeras such as BCR-ABL1 and TPM3-NTRK1 have been shown to be effective targets. Due to these past successes, there exists an opportunity to identify new chimeras as therapeutic targets. While modern chimera prediction software allows for the fast and accurate identification of chimeric RNAs from RNA sequencing data, investigations separating therapeutically relevant transcripts from “transcriptional noise” remain lacking. In this study we performed an in-silico functional screen of chimeras across cell lines representing a wide variety of cancers. We used state-of-the-art chimeric RNA prediction software to create a database of chimeric RNAs across 1017 cell lines from The Cancer Cell Line Encyclopedia. For each chimera we assessed factors such as frequency, recurrence, breakpoint coordinates, frame, and coding potential. To identify specific functional transcripts, we integrated publicly available shRNA knockdown data with our predictions and developed an in-silico functional analysis pipeline comparing differential knockdown effects between chimera and corresponding parental transcripts. For each transcript we assessed the average fold change of chimera-mapping probes, a function score defined as the difference between the fold change of chimera-mapping and parent-specific probes, and a p-value assessing the confidence of our functional score. From our initial screen of 127,819 transcripts, we identified 1088 high-confidence functional chimeras. We successfully identified nearly all known functional chimeras screened including PAX3-FOXO1, EWSR1-FLI1, and TCF3-PBX1. We also identified previously unknown chimeras that that we predict have a function in cell growth and proliferation. Overall, the results of our study reveal a new landscape of functionally relevant chimeric transcripts in cancer. Follow-up studies can further investigate these transcripts to determine their potential as therapeutic targets. Citation Format: Samir Lalani, Sehajroop Gadh, Justin Elfman, Sandeep Singh, Hui Li. Differential dependency mapping of chimeric RNAs across cancer reveals a new landscape of functional fusion transcripts [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 4351.