Search

Table S4

AACR GENIE Data Guide

Supplemental Methods. Supplemental Table 1: ââ,¬â€¹Genomic Data Characterization by Center. Supplemental Table 2: ââ,¬â€¹Gene Panels Submitted by Each Center. Figure S1: Number of putative germline SNPs per sample, before and after uniform germline filtering. Figure S2ââ,¬â€¹. Distribution of total somatic mutation burden per sample stratified by sequencing panel. Figure S3: ââ,¬â€¹Log-scale comparison of mutation frequencies at hotspot sites between GENIE (data aggregated from all sequencing panels) and cancerhotspots.org (CHS) using a binomial test. Figure S4:ââ,¬â€¹ Comparison of mutation frequencies at hotspot sites in each GENIE sequencing panel with cancerhotspots.org (CHS) using a binomial test.

PDF file, 156K, NFκB-regulated and inflammation-related genes used for gene expression analysis.

PDF file, 1402K, Differential expression of RelA and DAB2 proteins in secretory FTE cells.

PDF file, 98K.

PDF file, 27K, DAB2 and BRCA1 enhance dexamethasone-induced GR transactivation activity in pancreatic and breast cancer cell lines.

PDF file, 25K, Impact of DAB2 and BRCA1 on TNFα-induced BIRC3 expression.

PDF file, 59K, NFκB-regulated and inflammation-related genes differentially expressed in FTEb(S) samples (n=4) relative to remaining luteal phase FTE samples (n=8).

PDF file, 43K, NFκB-regulated and inflammation-related genes differentially expressed in FTEb(S) samples (n=4) relative to remaining FTE samples (n=20).

Purpose: To assess inflammation-related gene expression in nonmalignant fallopian tube epithelium (FTE) from BRCA1/2 mutation carriers and control patients obtained during the luteal and follicular phase, and to determine the impact of BRCA1 and disabled homolog 2 (DAB2) on NF-κB–mediated proinflammatory signaling.Experimental Design: A list of inflammation-related and NF-κB–responsive genes was compiled through gene set enrichment and PubMed database search, corresponding probes identified, and unpaired t tests conducted to identify differentially expressed genes in previously profiled FTE samples. ES2 and A549 cells were cotransfected with DAB2- or BRCA1-targeting siRNA and an NF-κB–responsive luciferase reporter, treated with TNF-α and luciferase activity determined. To determine whether DAB2 or BRCA1 alters mRNA expression of NF-κB target genes, cells were transfected with siRNA, treated with TNF-α, and harvested for total RNA extraction and quantitative real-time PCR.Results: A subset of BRCA1-mutated luteal phase samples previously found to group with adnexal high–grade serous carcinomas (HGSCs) differentially expressed 124 inflammation–associated probesets relative to remaining FTE samples. These samples also differentially expressed 264 probes relative to other luteal phase samples exposed to the same postovulatory environment. Both BRCA1- and DAB2–targeting siRNA increased TNF-α-induced NF-κB activity and mRNA expression of NF-κB–dependent target gene SOD2 relative to nontargeting siRNA, suggesting that both proteins repress proinflammatory signaling.Conclusions: These data provide evidence of elevated proinflammatory signaling in a subset of BRCA1-mutated luteal phase FTE, consistent with an altered response to ovulation-associated cytokines. Furthermore, both BRCA1 and DAB2 affect NF-κB activity, indicating a novel link between BRCA mutation status, ovulation, and predisposition to HGSC. Clin Cancer Res; 18(16); 4334–44. ©2012 AACR.

Supplementary Figure 3 from Plasminogen Activator uPA Is a Direct Transcriptional Target of the JAG1-Notch Receptor Signaling Pathway in Breast Cancer

Supplementary Table 3 from Plasminogen Activator uPA Is a Direct Transcriptional Target of the JAG1-Notch Receptor Signaling Pathway in Breast Cancer

Supplementary Table 4 from Plasminogen Activator uPA Is a Direct Transcriptional Target of the JAG1-Notch Receptor Signaling Pathway in Breast Cancer

Supplementary Figure 2 from Plasminogen Activator uPA Is a Direct Transcriptional Target of the JAG1-Notch Receptor Signaling Pathway in Breast Cancer

Supplementary Table 2 from Plasminogen Activator uPA Is a Direct Transcriptional Target of the JAG1-Notch Receptor Signaling Pathway in Breast Cancer

Supplementary Table 1 from Plasminogen Activator uPA Is a Direct Transcriptional Target of the JAG1-Notch Receptor Signaling Pathway in Breast Cancer

Supplementary Figure Legends 1-3, Methods from Plasminogen Activator uPA Is a Direct Transcriptional Target of the JAG1-Notch Receptor Signaling Pathway in Breast Cancer

Background Immense volumes of personal health information (PHI) are required to realize the anticipated benefits of artificial intelligence in clinical medicine. To maintain public trust in medical research, consent policies must evolve to reflect contemporary patient preferences. Methods Patients were invited to complete a 27-item survey focusing on: (a) broad versus specific consent; (b) opt-in versus opt-out approaches; (c) comfort level sharing with different recipients; (d) attitudes towards commercialization; and (e) options to track PHI use and study results. Results 222 participants were included in the analysis; 83% were comfortable sharing PHI with researchers at their own hospital, although younger patients (≤ 49 years) were more uncomfortable than older patients (50 + years; 13% versus 2% uncomfortable, p Conclusions While most patients were willing to share their PHI with researchers within their own institution, many preferred a transparent and reciprocal consent process. These data also suggest a generational shift, wherein younger patients preferred more specific consent options. Modernizing consent policies to reflect increased autonomy is crucial in fostering sustained public engagement with medical research.

Up to 20% of high-grade serous ovarian carcinomas (HGSOC) are hereditary; however, historical uptake of genetic testing is low. We used a unique combination of approaches to identify women in Ontario, Canada, with a first-degree relative (FDR) who died from HGSOC without prior genetic testing, and offer them multi-gene panel testing.From May 2015-Sept 2019, genetic counseling and testing was provided to eligible participants. Two recruitment strategies were employed, including self-identification in response to an outreach campaign and direct targeting of FDRs of deceased HGSOC patients treated at our institution. The rate of pathogenic variants (PV) in established/potential ovarian cancer risk genes and the benefits/challenges of each approach were assessed.A total of 564 women enrolled in response to our outreach campaign (n = 473) or direct recruitment (n = 91). Mean age at consent was 52 years and 96% did not meet provincial testing criteria. Genetic results were provided to 528 individuals from 458 families. The rate of PVs in ovarian cancer risk genes was highest when FDRs were diagnosed with HGSOC60 years (9.4% vs. 3.9% ≥ 60y, p = 0.0160). Participants in the outreach vs. direct recruitment cohort had a similar rate of PVs; however, uptake of genetic testing (97% vs. 89%; p = 0.0036) and study completion (95% vs. 87%; p = 0.0062) rates were higher in the former. Eleven participants with pathogenic variants have completed risk-reducing gynecologic surgery, with one stage I HGSOC and two breast cancers identified.Overall PV rates in this large cohort were lower than expected; however, we provide evidence that genetic testing criteria in Ontario should include individuals with a deceased FDR diagnosed with HGSOC60 years of age.

Background Rapid ethical access to personal health information (PHI) to support research is extremely important during pandemics, yet little is known regarding patient preferences for consent during such crises. This follow-up study sought to ascertain whether there were differences in consent preferences between pre-pandemic times compared to during Wave 1 of the COVID-19 global pandemic, and to better understand the reasons behind these preferences. Methods A total of 183 patients in the pandemic cohort completed the survey via email, and responses were compared to the distinct pre-pandemic cohort (n = 222); all were patients of a large Canadian cancer center. The survey covered (a) broad versus study-specific consent; (b) opt-in versus opt-out contact approach; (c) levels of comfort sharing with different recipients; (d) perceptions of commercialization; and (e) options to track use of information and be notified of results. Four focus groups (n = 12) were subsequently conducted to elucidate reasons motivating dominant preferences. Results Patients in the pandemic cohort were significantly more comfortable with sharing all information and biological samples (90% vs. 79%, p = 0.009), sharing information with the health care institution (97% vs. 83%, p p . 53%, p p = 0.022) and internationally (48% vs. 39%, p = 0.024) compared to the pre-pandemic cohort. Discomfort with sharing information with commercial companies remained unchanged between the two cohorts (50% vs. 51% uncomfortable, p = 0.58). Significantly more pandemic cohort patients expressed a wish to track use of PHI (75% vs. 61%, p = 0.007), and to be notified of results (83% vs. 70%, p = 0.012). Thematic analysis uncovered that transparency was strongly desired on outside PHI use, particularly when commercialization was involved. Conclusions In pandemic times, patients were more comfortable sharing information with all parties, except with commercial entities, where levels of discomfort (~ 50%) remained unchanged. Focus groups identified that the ability to track and receive results of studies using one’s PHI is an important way to reduce discomfort and increase trust. These findings meaningfully inform wider discussions on the use of personal health information for research during global crises.

BACKGROUND: Patients treated with neoadjuvant chemotherapy (NACT) for advanced high-grade serous ovarian carcinoma (HGSC) have a higher rate and shorter time to platinum-resistant recurrence compared to patients treated with primary cytoreductive surgery (PCS) and adjuvant chemotherapy. The purpose of this study is to determine the impact of NACT on somatic mutation status in platinum-sensitive and resistant HGSC. Patients with advanced HGSC who had a documented response to platinum-based NACT, a banked blood sample, and a banked tumor sample before and after NACT were identified. Whole exome and/or targeted deep sequencing was performed in matched normal and pre/post-NACT tumor samples from 3 platinum-resistant and 2 platinum-sensitive patients to identify somatic non-synonymous mutations at each time point.RESULTS: When comparing exonic non-synonymous mutations in pre-NACT and post-NACT samples from the same patient, an average of 41% (1-68%) of genes were mutated at both time points. There were no trends detected in the mutational burden following exposure to NACT in platinum-resistant vs. platinum-sensitive cases. The majority of mutated genes were unique to each case. We identified several genes that were commonly mutated in pre-NACT samples specific to platinum-resistant (CSPG4, SLC35G5, TUBA3D) or sensitive (CYP2D6, NUTM1, DNAH5) cases. Four mutated genes emerged exclusively in the platinum-resistant cases (ADGRV1, MUC17, MUC20, PAK2) following NACT.CONCLUSIONS: Patients with advanced HGSC present with significant intra-tumor heterogeneity. NACT significantly impacts the somatic mutation status irrespective of the time to recurrence. The mutated genes detected in chemo-naive pre-NACT tumor samples from either resistant or sensitive cases could potentially have a role in the prediction of chemotherapy response in patients scheduled to receive NACT; larger studies are required to further validate these genes.

Leukemia stem cells (LSCs) are linked to relapse in acute myeloid leukemia (AML). The LSC17 gene expression score robustly captures LSC stemness properties in AML and can be used to predict survival outcomes and response to therapy, enabling risk-adapted, upfront treatment approaches. The LSC17 score was developed and validated in a research setting. To enable widespread use of the LSC17 score in clinical decision making, we established a laboratory-developed test (LDT) for the LSC17 score that can be deployed broadly in clinical molecular diagnostic laboratories. We extensively validated the LSC17 LDT in a College of American Pathologists/Clinical Laboratory Improvements Act (CAP/CLIA)-certified laboratory, determining specimen requirements, a synthetic control, and performance parameters for the assay. Importantly, we correlated values from the LSC17 LDT to clinical outcome in a reference cohort of patients with AML, establishing a median assay value that can be used for clinical risk stratification of individual patients with newly diagnosed AML. The assay was established in a second independent CAP/CLIA-certified laboratory, and its technical performance was validated using an independent cohort of patient samples, demonstrating that the LSC17 LDT can be readily implemented in other settings. This study enables the clinical use of the LSC17 score for upfront risk-adapted management of patients with AML.

A common approach in clinical diagnostic laboratories to variant assessment from tumor molecular profiling is sequencing of genomic DNA extracted from both tumor (somatic) and normal (germline) tissue, with subsequent variant comparison to identify true somatic variants with potential impact on patient treatment or prognosis. However, challenges exist in paired tumor-normal testing, including increased cost of dual sample testing and identification of germline cancer predisposing variants. Alternatively, somatic variants can be identified by in silico tumor-only variant filtration precluding the need for matched normal testing. The barrier to tumor-only variant filtration is defining a reliable approach, with high sensitivity and specificity to identify somatic variants. In this study, we used retrospective data sets from paired tumor-normal samples tested on small (48 gene) and large (555 gene) targeted next-generation sequencing panels, to model algorithms for tumor-only variants classification. The optimal algorithm required an ordinal filtering approach using information from variant population databases (1000 Genomes Phase 3, ESP6500, ExAC), clinical mutation databases (ClinVar), and information on recurring clinically relevant somatic variants. Overall the tumor-only variant filtration strategy described in this study can define clinically relevant somatic variants from tumor-only analysis with sensitivity of 97% to 99% and specificity of 87% to 94%, and with significant potential utility for clinical laboratories implementing tumor-only molecular profiling.

γ-Aminobutyric acid (GABA) administration has been shown to increase β-cell mass, leading to a reversal of type 1 diabetes in mice. Whether GABA has any effect on β cells of healthy and prediabetic/glucose-intolerant obese mice remains unknown. In the present study, we show that oral GABA administration ( ad libitum) to mice indeed increased pancreatic β-cell mass, which led to a modest enhancement in insulin secretion and glucose tolerance. However, GABA treatment did not further increase insulin-positive islet area in high fat diet-fed mice and was unable to prevent or reverse glucose intolerance and insulin resistance. Mechanistically, whether in vivo or in vitro, GABA treatment increased β-cell proliferation. In vitro, the effect was shown to be mediated via the GABAA receptor. Single-cell RNA sequencing analysis revealed that GABA preferentially up-regulated pathways linked to β-cell proliferation and simultaneously down-regulated those networks required for other processes, including insulin biosynthesis and metabolism. Interestingly, single-cell differential expression analysis revealed GABA treatment gave rise to a distinct subpopulation of β cells with a unique transcriptional signature, including urocortin 3 ( ucn3), wnt4, and hepacam2. Taken together, this study provides new mechanistic insight into the proliferative nature of GABA but suggests that β-cell compensation associated with prediabetes overlaps with, and negates, its proliferative effects.-Untereiner, A., Abdo, S., Bhattacharjee, A., Gohil, H., Pourasgari, F., Ibeh, N., Lai, M., Batchuluun, B., Wong, A., Khuu, N., Liu, Y., Al Rijjal, D., Winegarden, N., Virtanen, C., Orser, B. A., Cabrera, O., Varga, G., Rocheleau, J., Dai, F. F., Wheeler, M. B. GABA promotes β-cell proliferation, but does not overcome impaired glucose homeostasis associated with diet-induced obesity.

Rapid expansions of bioinformatics and computational biology have broadened the collection and use of -omics data including genomic, transcriptomic, methylomic and a myriad of other health data types, in the clinic and the laboratory. Both clinical and research uses of such data require co-analysis with large datasets, for which participant privacy and the need for data custodian controls must remain paramount. This is particularly challenging in multi-jurisdictional settings, such as Canada, where health privacy and security requirements are often heterogeneous. Data federation presents a solution to this, allowing for integration and analysis of large datasets from various sites while abiding by local policies.The Canadian Distributed Infrastructure for Genomics platform (CanDIG) enables federated querying and analysis of -omics and health data while keeping that data local and under local control. It builds upon existing infrastructures to connect five health and research institutions across Canada, relies heavily on standards and tooling brought together by the Global Alliance for Genomics and Health (GA4GH), implements a clear division of responsibilities among its participants and adheres to international data sharing standards. Participating researchers and clinicians can therefore contribute to and quickly access a critical mass of -omics data across a national network in a manner that takes into account the multi-jurisdictional nature of our privacy and security policies. Through this, CanDIG gives medical and research communities the tools needed to use and analyze the ever-growing amount of -omics data available to them in order to improve our understanding and treatment of various conditions and diseases. CanDIG is being used to make genomic and phenotypic data available for querying across Canada as part of data sharing for five leading pan-Canadian projects including the Terry Fox Comprehensive Cancer Care Centre Consortium Network (TF4CN) and Terry Fox PRecision Oncology For Young peopLE (PROFYLE), and making data from provincial projects such as POG (Personalized Onco- Genomics) more widely available.

Introduction Increasingly, the label “data trust” is being applied to repeatable mechanisms or approaches to sharing data in a timely, fair, safe, and equitable way. However, there is an absence of practical guidance regarding how to establish and operate a data trust. Aim and approach In December 2019, the Canadian Institute for Health Information and the Vector Institute for Artificial Intelligence convened a working meeting of 19 people representing 15 Canadian organizations/initiatives involved in data sharing, most of which focus on public sector health data. The objective was to identify essential requirements for the establishment and operation of data trusts in the Canadian context. Preliminary requirements were discussed during the meeting and then refined as authors contributed to this manuscript. Results Twelve minimum specification requirements (“min specs”) for data trusts were identified. The foundational min spec is that data trusts must meet all legal requirements, including legal authority to collect, hold or share data. In addition, there was agreement that data trusts must have (i) an accountable governing body to ensure that the data trust achieves its stated purpose and is transparent, (ii) comprehensive data management including clear processes and qualified individuals responsible for the collection, storage, access, disclosure and use of data, (iii) training and accountability requirements for all data users and (iv) ongoing public and stakeholder engagement. Conclusions Practical guidance for the establishment and operation of data trusts was articulated in the form of 12 min specs requirements. The 12 min specs are a starting point. Future work to refine and strengthen them with members of the public, companies, and additional research data stakeholders from within and outside of Canada, is recommended.

Key Points Question Can virtual care (VC) be rapidly implemented across a tertiary center during the coronavirus disease 2019 (COVID-19) pandemic, and what are service capacity and quality outcomes? Findings This cohort study of 22 085 VC visits at a single cancer center suggests feasibility of an agile service design process for implementation of VC at scale. This approach preserved outpatient caseloads and maintained care quality in all 6 care-quality domains of care quality laid out by the Institute of Medicine while rendering high patient and practitioner satisfaction. Meaning These data support the value proposition of VC to safeguard system capacity, while minimizing the disruption to patient care during a pandemic., Importance The coronavirus disease 2019 (COVID-19) pandemic has burdened health care resources and disrupted care of patients with cancer. Virtual care (VC) represents a potential solution. However, few quantitative data support its rapid implementation and positive associations with service capacity and quality. Objective To examine the outcomes of a cancer center–wide virtual care program in response to the COVID-19 pandemic. Design, Setting, and Participants This cohort study applied a hospitalwide agile service design to map gaps and develop a customized digital solution to enable at-scale VC across a publicly funded comprehensive cancer center. Data were collected from a high-volume cancer center in Ontario, Canada, from March 23 to May 22, 2020. Main Outcomes and Measures Outcome measures were care delivery volumes, quality of care, patient and practitioner experiences, and cost savings to patients. Results The VC solution was developed and launched 12 days after the declaration of the COVID-19 pandemic. A total of 22 085 VC visits (mean, 514 visits per day) were conducted, comprising 68.4% (range, 18.8%-100%) of daily visits compared with 0.8% before launch (P, This cohort study of a publicly funded comprehensive cancer center examines the outcomes of a cancer center–wide virtual care program in response to the COVID-19 pandemic.

IntroductionThere is widespread enthusiasm to improve health through the application of artificial intelligence and machine learning (AI/ML) methods to large population-level health datasets. Achieving this may require successful collaboration between institutions as well as between computer scientists (CS), machine learning researchers (MLR) and health service researchers (HSR). The objective is to describe lessons learned in creating the Health Artificial Data and Analysis Platform (HAIDAP) in Ontario, Canada. Objectives and ApproachA partnership between a HSR institute (ICES), an AI/ML institute (Vector) and a high-performance computing center (HPC4H) was initiated in 2017 to enable the application of AI/ML methods to population-level health data for the province of Ontario (population 14M). The HAIDAP was launched in 2019. We describe lessons learned (and being learned) following the HAIDAP’s launch. ResultsMajor learnings include: 1)importance of institutional partnerships and alignment with institutional strategies; 2)potential of joint institutional risk-sharing models; 3)need for scientific collaborations bridging disciplines around joint research projects; 4)sensitivity to different scientific cultures (e.g., academic prestige of conference proceedings for MLR vs journal publications for HSR; traditional statistical vs. ML model assumptions); 5)differences in research timeline expectations; 6)different experience with and expectations for access to de-identified routinely collected data (e.g., need for research ethics committee project approvals and privacy impact assessments); 7)developing data access models that enable greater flexibility (e.g. importing code or using open source tools); 8) broadening data access models to allow modern high-dimensional exploratory data analysis; 9)obtaining support of information/privacy regulator; 10)hardware is (relatively) easy part compared to other success factors. Conclusion / ImplicationsThe HAIDAP has enabled multi-disciplinary collaborations and novel AI/ML research of Ontario’s population-level health data. Collectively we have learned that additional effort is required to develop systems and processes enabling more efficient access to data and analytic tools for the analysis of administrative health data.

Introduction: Increasingly, the label data trust is being applied to repeatable mechanisms or approaches to sharing data in a timely, fair, safe and equitable way. However, there is a gap in terms of practical guidance about how to establish and operate a data trust. Aim and Approach: In December 2019, the Canadian Institute for Health Information and the Vector Institute for Artificial Intelligence convened a working meeting of 19 people representing 15 Canadian organizations/initiatives involved in data sharing, most of which focus on public sector health data. The objective was to identify essential requirements for the establishment and operation of data trusts. Preliminary findings were presented during the meeting then refined as participants and co-authors identified relevant literature and contributed to this manuscript. Results: Twelve (12) minimum specification requirements (min specs) for data trusts were identified. The foundational min spec is that data trusts must meet all legal requirements, including legal authority to collect, hold or share data. In addition, there was agreement that data trusts must have (i) an accountable governing body which ensures the data trust advances its stated purpose and is transparent, (ii) comprehensive data management including responsible parties and clear processes for the collection, storage, access, disclosure and use of data, (iii) training and accountability requirements for all data users and (iv) ongoing public and stakeholder engagement. Conclusion / Implications: Based on a review of the literature and advice from participants from 15 Canadian organizations/initiatives, practical guidance in the form of twelve min specs for data trusts were agreed on. Public engagement and continued exchange of insights and experience is recommended on this evolving topic., Comment: 17 pages including references, 1 text box

Cancer is a genetic disease resulting from germline or somatic genetic aberrations. Rapid progress in the field of genomics in recent years is allowing for increased characterization and understanding of the various forms of the disease. The Ontario-wide Cancer Targeted Nucleic Acid Evaluation (octane) clinical trial, open at cancer centres across Ontario, aims to increase access to genomic sequencing of tumours and to facilitate the collection of clinical data related to enrolled patients and their clinical outcomes. The study is designed to assess the clinical utility of next-generation sequencing (ngs) in cancer patient care, including enhancement of treatment options available to patients. A core aim of the study is to encourage collaboration between cancer hospitals within Ontario while also increasing international collaboration in terms of sharing the newly generated data. The single-payer provincial health care system in Ontario provides a unique opportunity to develop a province-wide registry of ngs testing and a repository of genomically characterized, clinically annotated samples. It also provides an important opportunity to use province-wide real-world data to evaluate outcomes and the cost of ngs for patients with advanced cancer. The octane study is attempting to translate knowledge to help deliver precision oncology in a Canadian environment. In this article, we discuss the background to the study and its implementation, current status, and future directions.

Background and rationale There is widespread enthusiasm to improve health through the application of artificial intelligence and machine learning (AI/ML) methods to large population-level health datasets. Achieving this may require successful collaboration between institutions as well as between computer scientists (CS), machine learning researchers (MLR) and health service researchers (HSR). Main Aim Describe lessons learned in creating the Health Artificial Data and Analysis Platform (HAIDAP) in Ontario, Canada. Methods/Approach A partnership between a HSR institute (ICES), an AI/ML institute (Vector) and a high-performance computing center (HPC4H) was initiated in 2017 to enable the application of AI/ML methods to population-level health data for the province of Ontario (population 14M). We describe lessons learned (and being learned) following the HAIDAP’s launch. Results The HAIDAP was launched in 2019. Major learnings include: 1) importance of institutional partnerships and alignment with institutional strategies; 2) potential of joint institutional risk-sharing models; 3) need for scientific collaborations bridging disciplines around joint research projects; 4) sensitivity to different scientific cultures (e.g., academic prestige of conference proceedings for MLR vs journal publications for HSR; traditional statistical vs. ML model assumptions); 5) differences in research timeline expectations; 6) different experience with and expectations for access to de-identified routinely collected data (e.g., need for research ethics committee project approvals and privacy impact assessments); 7) developing data access models that enable greater flexibility (e.g., importing code or using open source tools); 8) broadening data access models to allow modern high-dimensional exploratory data analysis; 9) obtaining support of information/privacy regulator; 10) the hardware is the (relatively) easy part compared to other success factors. Conclusion The HAIDAP has enabled multi-disciplinary collaborations and novel AI/ML research of Ontario’s population-level health data. Collectively we have learned that additional effort is required to develop systems and processes enabling more efficient access to data and analytic tools for the analysis of administrative health data.

Objectives To identify differentially expressed genes between relapsed and non-relapsed clinical stage I testicular germ cell tumours (TGCTs). Materials and methods We reviewed patients with clinical stage I non-seminoma and seminoma from an institutional database (2000-2012) who were managed by active surveillance. Patients with non-relapsed non-seminoma and non-relapsed seminoma were defined as being relapse-free after 2 and 3 years of surveillance, respectively. RNA extraction and gene expression analysis was performed on archival primary tumour samples and gene-set enrichment analysis (GSEA) was conducted in order to identify differentiating biological pathways. Results A total of 57 patients (relapsed non-seminoma, n = 12; relapsed seminoma, n =15; non-relapsed non-seminoma, n = 15; non-relapsed seminoma, n = 15) were identified, with a median (range) relapse time of 5.6 (2.5-18.1) and 19.3 (4.7-65.3) months in the relapsed non-seminoma and relapsed seminoma cohorts, respectively. A total of 1 039 differentially expressed genes were identified that separated relapsed and non-relapsed groups. In patients with relapse, GSEA revealed enrichment in pathways associated with differentiation, such as skeletal development (i.e. FGFR1, BMP4, GLI2, SPARC, COL2A1), tissue (i.e. BMP4, SPARC, COL13A1) and bone remodelling (i.e. CARTPT, GLI2, MGP). A discriminative gene expression profile between relapsed and non-relapsed cases was discovered when combining non-seminoma and seminoma samples using 10- and 30-probe signatures; however, this profile was not observed in the seminoma and non-seminoma cohorts individually. Conclusion A discriminating signature for relapsed disease was identified for clinical stage I TGCT that we were not able to identify by histology alone. Further validation is required to determine if this signature provides independent prognostic information to standard pathological risk factors.

We present the Canadian Distributed Infrastructure for Genomics (CanDIG) platform, which enables federated querying and analysis of human genomics and linked biomedical data. CanDIG leverages the standards and frameworks of the Global Alliance for Genomics and Health (GA4GH) and currently hosts data for five pan-Canadian projects. We describe CanDIG’s key design decisions and features as a guide for other federated data systems.