Wang, Chen

Abstract Background: Benign breast disease (BBD) is a risk factor for breast cancer (BC); however, little is known about the genetic alterations present at the time of BBD diagnosis, and how these relate to risk of incident BC. Methods: A subset of a long-term BBD cohort was selected to examine DNA variation across three BBD groups (42 future estrogen receptor-positive (ER+) BC, 36 future estrogen receptor-negative (ER-) BC and 42 controls cancer-free for at least 16 years post-BBD). DNA extracted from archival formalin fixed, paraffin-embedded (FFPE) tissue blocks was tested and analyzed for presence of DNA alterations using a targeted panel of 93 BC-associated genes. To address artifacts frequently observed in FFPE tissues (e.g., C>T changes), we applied three filtering strategies based on alternative allele frequencies and nucleotide substitution context. Gene-level associations were performed using two burden tests, SKAT-O and logistic regression, and adjusted for clinical and technical covariates including epithelial percentage, histologic impression, patient age and year of BBD biopsy. Results: After filtering, the variant frequency of SNPs in our samples was highly consistent with population allele frequencies reported in 1KG/ExAC (0.986, p <1e-16). Ten genes were found to be associated with later cancer status by four of 12 association methods, with nominal p-values <0.05: MED12, MSH2, BRIP1, PMS1, GATA3, MUC16, FAM175A, EXT2, MLH1 and TGFB1. Additionally, analysis revealed 11 and 19 gene-level associations with p-values <0.05 and OR<1 for all cancer cases and ER- cases, respectively. Comparing between the three case groups, BBD ER+ cases were closer to controls in mutation profile, while BBD ER- cases were distinct. Notably, the variant burden was significantly higher in controls than in either ER+ or ER- cases. CD45 expression was associated with mutational burden (p=0.003).Conclusions: Somatic mutations were more frequent in benign breast tissue from women who did not develop cancer, opening questions of clonal diversity or immune-mediated restraint on future cancer development. CD45 expression was positively associated with mutational burden, most strongly in controls. Further studies in both normal and premalignant tissues are needed to better understand the role of somatic gene mutations and their contribution to future cancer development.

Abstract Background: Next-generation sequencing provides comprehensive information about individuals’ genetic makeup and is commonplace in precision oncology practice. Due to the heterogeneity of individual patient’s disease conditions and treatment journeys, not all targeted therapies were initiated despite actionable mutations. To better understand and support the clinical decision-making process in precision oncology, there is a need to examine real-world associations of patients’ genetic information and treatment choice.Methods: To fill the gap of insufficient use of real-world data (RWD) in electronic health records (EHRs), we generated a single Resource Description Framework (RDF) resource, called PO2RDF (precision oncology to RDF) by integrating information regarding gene, variant, disease, and drug from genetic reports and EHRs. Results: There are total 2,309,014 triples contained in the PO2RDF. Among them 32,815 triples are related to Gene, 34,695 triples are related to Variant, 8,787 triples are related to Disease, 26,154 triples are related to Drug. We performed one use case analysis to demonstrate the usability of the PO2RDF: we examined real-world associations between EGFR mutations and targeted therapies to confirm existing knowledge and detect off-label use. Conclusions: In conclusion, our work proposed to use RDF to organize and distribute clinical RWD that is otherwise inaccessible externally. Our work serves as a pilot study that will lead to new clinical applications and could ultimately stimulate progress in the field of precision oncology.

Abstract Background: Traditionally, mutational burden and mutational signatures have been assessed by tumor-normal pair DNA sequencing. The requirement of having both normal and tumor samples is not always feasible from a clinical perspective, and led us to investigate the efficacy of using RNA sequencing of only the tumor sample to determine the mutational burden and signatures, and subsequently molecular cause of the cancer. The potential advantages include reducing the cost of testing, and simultaneously providing information on the gene expression profile and gene fusions present in the tumor. Results: In this study, we devised supervised and unsupervised learning methods to determine mutational signatures from tumor RNA-seq data. As applications, we applied the methods to a training set of 587 TCGA uterine cancer RNA-seq samples, and examined in an independent testing set of 521 TCGA colorectal cancer RNA-seq samples. Both diseases are known associated with microsatellite instable high (MSI-H) and driver defects in DNA polymerase ɛ (POLɛ). From RNA-seq called variants, we found majority (>95%) are likely germline variants, leading to C>T enriched germline variants (dbSNP) widely applicable in tumor and normal RNA-seq samples. We found significant associations between RNA-derived mutational burdens and MSI/POLɛ status, and insignificant relationship between RNA-seq total coverage and derived mutational burdens. Additionally we found that over 80% of variants could be explained by using the COSMIC mutational signature-5, -6 and -10, which are implicated in natural aging, MSI-H, and POLɛ, respectively. For classifying tumor type, within UCEC we achieved a recall of 0.56 and 0.78, and specificity of 0.66 and 0.99 for MSI and POLɛ respectively. By applying learnt RNA signatures from UCEC to COAD, we were able to improve our classification of both MSI and POLɛ. Conclusions: Taken together, our work provides a novel method to detect RNA-seq derived mutational signatures with effective procedures to remove likely germline variants. It can leads to accurate classification of underlying driving mechanisms of DNA damage deficiency.

Abstract Background: Next-generation sequencing provides comprehensive information about individuals’ genetic makeup and is commonplace in oncology clinical practice. However, the utility of genetic information in clinical decision-making process has not been examined extensively from a real-world, data-driven perspective. Through mining real-world data (RWD) from clinical notes, we could extract patients’ genetic information and further associate treatment decisions with genetic information.Methods: We proposed a real-world evidence (RWE) study framework that incorporates context-based natural language processing (NLP) methods and data quality examination before final association analysis. The framework was demonstrated in a Foundation-tested women cancer cohort (N=196). Upon retrieval of patients’ genetic information using NLP system, we assessed the completeness of genetic data captured in unstructured clinical notes according to a genetic data-model. We examined the distribution of different topics regarding BRCA1/2 throughout patients’ treatment process, and then analyzed the association between BRCA1/2 mutation status and the discussion/prescription of targeted therapy. Results: We identified seven topics in clinical context of genetic mentions including: Information, Evaluation, Insurance, Order, Negative, Positive, and Variants of unknown significance (VUS). Our rule-based system achieved a precision of 0.87, recall of 0.93 and F-measure of 0.91. Our machine learning system achieved a precision of 0.901, recall of 0.899 and F-measure of 0.9 for four-topic classification and a precision of 0.833, recall of 0.823 and F-measure of 0.82 for seven-topic classification. We found in result-containing sentences, capture of BRCA1/2 mutation information was 75%, but detailed variant information (e.g. variant types) is largely missing. Using cleaned RWD, significant associations were found between BRCA1/2 positive mutation and targeted therapies.Conclusions: In conclusion, we demonstrated a framework to generate RWE using RWD from different clinical sources. Rule-based NLP system achieved the best performance for resolving contextual variability when extracting RWD from unstructured clinical notes. Data quality issues such as incompleteness and discrepancies exist thus manual data cleaning is needed before further analysis can be performed. Finally, we were able to use cleaned RWD to evaluate real-world utility of genetic information to initiate a prescription of targeted therapy.

Abstract Background: Traditionally, mutational burden and mutational signatures have been assessed by tumor-normal pair DNA sequencing. The requirement of having both normal and tumor samples is not always feasible from a clinical perspective, and led us to investigate the efficacy of using RNA sequencing of only the tumor sample to determine the mutational burden and signatures, and subsequently molecular cause of the cancer. The potential advantages include reducing the cost of testing, and simultaneously providing information on the gene expression profile and gene fusions present in the tumor. Results: In this study, we devised supervised and unsupervised learning methods to determine mutational signatures from tumor RNA-seq data. As applications, we applied the methods to a training set of 587 TCGA uterine cancer RNA-seq samples, and examined in an independent testing set of 521 TCGA colorectal cancer RNA-seq samples. Both diseases are known associated with microsatellite instable high (MSI-H) and driver defects in DNA polymerase ɛ (POLɛ).From RNA-seq called variants, we found majority (>95%) are likely germline variants, leading to C>T enriched germline variants (dbSNP) widely applicable in tumor and normal RNA-seq samples. We found significant associations between RNA-derived mutational burdens and MSI/POLɛ status, and insignificant relationship between RNA-seq total coverage and derived mutational burdens. Additionally we found that over 80% of variants could be explained by using the COSMIC mutational signature-5, -6 and -10, which are implicated in natural aging, MSI-H, and POLɛ, respectively. For classifying tumor type, within UCEC we achieved a recall of 0.56 and 0.78, and specificity of 0.66 and 0.99 for MSI and POLɛ respectively. By applying learnt RNA signatures from UCEC to COAD, we were able to improve our classification of both MSI and POLɛ. Conclusions: Taken together, our work provides a novel method to detect RNA-seq derived mutational signatures with effective procedures to remove likely germline variants. It can leads to accurate classification of underlying driving mechanisms of DNA damage deficiency.

Abstract Background : Next- generation sequencing provides comprehensive information about individuals’ genetic makeup and is commonplace in oncology clinical practice. However, the utility of genetic information in clinical decision-making process has not been examined extensively from a real-world, data-driven perspective. Through mining real-world data (RWD) from clinical notes, we could extract patients’ genetic information and further associate treatment decisions with genetic information. Methods : We proposed a real-world evidence (RWE) study framework that incorporates context-based natural language processing (NLP) methods and data quality examination before final association analysis. The framework was demonstrated in a Foundation-tested women cancer cohort (N=196). Upon retrieval of patients’ genetic information using NLP system, we assessed the completeness of genetic data captured in unstructured clinical notes according to a genetic data-model. We examined the distribution of different topics regarding BRCA1/2 throughout patients’ treatment process, and then analyzed the association between BRCA1/2 mutation status and the discussion/prescription of targeted therapy . Results: We identified seven topics in clinical context of genetic mentions including: Information, Evaluation, Insurance, Order, Negative, Positive, and Variants of unknown significance (VUS) . O ur rule-based system achieved a precision of 0.87, recall of 0.93 and F- measure of 0.91. Our machine learning system achieved a precision of 0.901, recall of 0.899 and F- measure of 0.9 for four-topic classification and a precision of 0.833, recall of 0.823 and F- measure of 0.82 for seven-topic classification. We found in result-containing sentences, capture of BRCA1/2 mutation information was 75%, but detailed variant information (e.g. variant types) is largely missing. Using cleaned RWD, significant associations were found between BRCA1/2 positive mutation and targeted therapies. Conclusions : In conclusion, we demonstrated a framework to generate RWE using RWD from different clinical sources. Rule-based NLP system achieved the best performance for resolving contextual variability when extracting RWD from unstructured clinical notes . Data quality issues such as incompleteness and discrepancies exist thus manual data cleaning is needed before further analysis can be performed. Finally, we were able to use cleaned RWD to evaluate real-world utility of genetic information to initiate a prescription of targeted therapy.

Abstract Introduction: Acquired resistance to platinum remains a major clinical problem in ovarian cancer since systemic therapies have limited efficacy in this setting. There is growing evidence that the tumor microenvironment (TME) contributes to such resistance, as well as the pathogenesis of the disease. The study of the TME is limited by technical challenges of separating the epithelial from the stromal component. To overcome this limitation, we performed RNA sequencing (RNA-Seq) for three pairs of sensitive and induced resistant patient-derived xenograft models since the epithelial transcripts originate from human tissue and the stroma component is murine. Methods: To determine which TME genes might be contributing to chemotherapy resistance, three platinum-sensitive ovarian cancer PDX lines were treated repeatedly with carboplatin/paclitaxel until resistance occurred. Primary platinum-sensitive and induced platinum resistant tumors were characterized histologically by H&E and Masson Trichrome staining. The three pairs of sensitive and resistant PDX models were then subjected to RNA-Seq. Raw sequencing data were processed through Xenome to classify the reads into epithelial carcinoma (human) or TME (mouse) through Mayo Clinic’s internal RNA-Seq pipeline for transcriptome quantification. Differential expression analysis between induced resistant and sensitive models was carried out for both carcinoma and TME components using edgeR paired analysis. Gene set enrichment analysis was used to identify overrepresented biologic functions for differentially expressed genes between the pairs. Top significantly varied, differentially expressed genes between the pairs were validated using species-specific qPCR. Results: Histologic analysis of the matched sensitive/resistant pairs revealed a marked increase of the stromal component in the resistant tumors evident by H&E and Masson trichrome staining. A total number of 95 and 312 genes were differentially expressed in the epithelial and TME component, respectively (FDR&lt;0.05; absolute log2 fold-change &gt;1.5). The stemness marker SOX2 was significantly upregulated in the human component of the resistant tumors as confirmed by qPCR. Genes involved in stroma remodeling were upregulated in the “murine” component of the resistant tumors, and upregulation was confirmed by qPCR (normalized to a murine housekeeper gene). Gene set enrichment analysis showed that the upregulated genes in TME of resistant models were enriched for cytoskeleton related functions, while the downregulated genes were enriched for extracellular matrix-related functions. Conclusion: A bioinformatics strategy able to map human and murine reads separately may identify resistance mechanisms attributed to the two tumor subcompartments and could lead to strategies to potentially target each separately or concurrently. Citation Format: Valentina Zanfagnin, Yuanhang Liu, Pritha Chanana, Xiaonan Pritha, Beverly Long, Joe Pathoulas, Brad Evans, Jaime Davila, Chen Wang, Saravut John Weroha. Transcriptional profiling of tumor stroma using ovarian cancer PDX models with induced platinum-resistance [abstract]. In: Proceedings of the AACR Special Conference on Advances in Ovarian Cancer Research; 2019 Sep 13-16, 2019; Atlanta, GA. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(13_Suppl):Abstract nr B18.

Abstract Background: Gene expression-based molecular subtypes of high-grade serous tubo-ovarian cancer (HGSOC) are distinguished by differential immune and stromal infiltration and may provide opportunities for targeted therapies. Integration of molecular subtypes into clinical trials has been hindered by inconsistent subtyping methodology. Methods: Adopting two independent approaches, we derived and internally validated algorithms for molecular subtype prediction from gene-expression array data in 1,650 tumors. We applied resulting models to assign labels to 3829 HGSOCs from the Ovarian Tumor Tissue Analysis (OTTA) consortium evaluated on NanoString. Using the labeled NanoString data, we developed, confirmed, and validated a minimal gene set, clinical-grade test and prediction tool. We also used the OTTA dataset to evaluate associations between molecular subtype, biologic, and clinical features. Findings: The locked-down test included a model with 55 genes that predicted HGSOC molecular subtype with &gt;95% accuracy. Subtype varied between primary and metastatic site taken at the time of primary surgery, and was significantly associated with age, stage, CD8+ lymphocyte infiltration, residual disease, and outcome. In multivariable models, molecular subtypes lose their prognostic significance in the presence of risk factors such as residual disease and BRCA1/2 germline mutations. Interpretation: We validated the Predictor of high-grade-serous Ovarian carcinoma molecular subTYPE, or PrOTYPE, following the Institute of Medicine guidelines for the development of omics-based tests. This simple-to-use, cost-effective, fully defined, and locked-down clinical-grade assay will facilitate molecular subtype stratification into clinical trial design. PrOTYPE will allow for objective assessment of HGSOC molecular subtype predictive value in precision medicine applications. Citation Format: Aline Talhouk, Joshy George, Chen Wang, Ellen Goode, Susan Ramus, Jennifer Doherty, David Bowtell, Michael Anglesio, OTTA Consortium. PrOTYPE (Predictor of high-grade-serous Ovarian carcinoma molecular subTYPE): The development and validation of a clinical-grade consensus classifier for the molecular subtypes of high-grade serous tubo-ovarian cancer [abstract]. In: Proceedings of the AACR Special Conference on Advances in Ovarian Cancer Research; 2019 Sep 13-16, 2019; Atlanta, GA. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(13_Suppl):Abstract nr A03.

Abstract Background: Next-generation sequencing provides comprehensive information about individuals’ genetic makeup and is commonplace in oncology clinical practice. However, the utility of genetic information in clinical decision-making process has not been examined extensively from a real-world, data-driven perspective. Through mining real-world data (RWD) from clinical notes, we could extract patients’ genetic information and further associate treatment decisions with genetic information.Methods: We proposed a real-world evidence (RWE) study framework that incorporates context-based natural language processing (NLP) methods and data quality examination before final association analysis. The framework was demonstrated on a Foundation-tested women cancer cohort (N=196). Upon retrieval of patients’ genetic information using NLP system, we assessed completeness of genetic data captured in unstructured clinical notes according a genetic data-model. We examined the distribution of different topics regarding BRCA1/2 throughout patients’ treatment process, and then analyzed the association between BRCA1/2 mutation status and the discussion/prescription of targeted therapy. Results: We identified seven topics in clinical context of genetic mentions including: Information, Evaluation, Insurance, Order, Negative, Positive, and Variants of unknown significance (VUS). Our rule-based system achieved a precision of 0.87, recall of 0.93 and F-measure of 0.91. Our machine learning system achieved a precision of 0.901, recall of 0.899 and F-measure of 0.9 for four-topic classification and a precision of 0.833, recall of 0.823 and F-measure of 0.82 for seven-topic classification. We found in result-containing sentences, capture of BRCA1/2 mutation information was 75%, but detailed variant information (e.g. variant types) is largely missing. Using cleaned RWD, significant associations were found between BRCA1/2 positive mutation and targeted therapies.Conclusions: In conclusion, we demonstrated a framework to generate RWE using RWD from different clinical sources. Rule-based NLP system achieved the best performance for resolving contextual variability when extracting RWD from unstructured clinical notes. Data quality issue such as incompleteness and discrepancies exist thus manual data cleaning is needed before further analysis can be performed. Finally, we were able to use cleaned RWD to evaluate real-world utility of genetic information to initiate prescription of targeted therapy.

Abstract The Hippo pathway effector Yes-associated protein (YAP) is localized to the nucleus and transcriptionally active in a number of tumor types, including a majority of human cholangiocarcinomas. YAP activity has been linked to chemotherapy resistance and has been shown to rescue KRAS and BRAF inhibition in RAS/RAF-driven cancers; however, the underlying mechanisms of YAP-mediated chemoresistance have yet to be elucidated. Herein, we report that the tyrosine phosphatase SHP2 directly regulates the activity of YAP by dephosphorylating pYAPY357 even in the setting of RAS/RAF mutations, and that diminished SHP2 phosphatase activity is associated with chemoresistance in cholangiocarcinomas. A screen for YAP-interacting tyrosine phosphatases identified SHP2, and characterization of cholangiocarcinomas cell lines demonstrated an inverse relationship between SHP2 levels and pYAPY357. Human sequencing data demonstrated lower SHP2 levels in cholangiocarcinomas tumors as compared with normal liver. Cell lines with low SHP2 expression and higher levels of pYAPY357 were resistant to gemcitabine and cisplatin. In cholangiocarcinomas cells with high levels of SHP2, pharmacologic inhibition or genetic deletion of SHP2 increased YAPY357 phosphorylation and expression of YAP target genes, including the antiapoptotic regulator MCL1, imparting resistance to gemcitabine and cisplatin. In vivo evaluation of chemotherapy sensitivity demonstrated significant resistance in xenografts with genetic deletion of SHP2, which could be overcome by utilizing an MCL1 inhibitor. Implications: These findings demonstrate a role for SHP2 in regulating YAP activity and chemosensitivity, and suggest that decreased phosphatase activity may be a mechanism of chemoresistance in cholangiocarcinoma via a MCL1-mediated mechanism.