No abstract

Selection of effective genes that accurately predict chemotherapy response could improve cancer outcomes. We compare optimized gene signatures for cisplatin, carboplatin, and oxaliplatin response in the same cell lines, and respectively validate each with cancer patient data. Supervised support vector machine learning was used to derive gene sets whose expression was related to cell line GI50 values by backwards feature selection with cross-validation. Specific genes and functional pathways distinguishing sensitive from resistant cell lines are identified by contrasting signatures obtained at extreme vs. median GI50 thresholds. Ensembles of gene signatures at different thresholds are combined to reduce dependence on specific GI50 values for predicting drug response. The most accurate models for each platin are: cisplatin: BARD1, BCL2, BCL2L1, CDKN2C, FAAP24, FEN1, MAP3K1, MAPK13, MAPK3, NFKB1, NFKB2, SLC22A5, SLC31A2, TLR4, TWIST1; carboplatin: AKT1, EIF3K, ERCC1, GNGT1, GSR, MTHFR, NEDD4L, NLRP1, NRAS, RAF1, SGK1, TIGD1, TP53, VEGFB, VEGFC; oxaliplatin: BRAF, FCGR2A, IGF1, MSH2, NAGK, NFE2L2, NQO1, PANK3, SLC47A1, SLCO1B1, UGT1A1. TCGA bladder, ovarian and colorectal cancer patients were used to test cisplatin, carboplatin and oxaliplatin signatures (respectively), resulting in 71.0%, 60.2% and 54.5% accuracy in predicting disease recurrence and 59%, 61% and 72% accuracy in predicting remission. One cisplatin signature predicted 100% of recurrence in non-smoking bladder cancer patients (57% disease-free; N=19), and 79% recurrence in smokers (62% disease-free; N=35). This approach should be adaptable to other studies of chemotherapy response, independent of drug or cancer types.

We present an improved approach for learning dependency parsers from treebank data. Our technique is based on two ideas for improving large margin training in the context of dependency parsing. First, we incorporate local constraints that enforce the correctness of each individual link, rather than just scoring the global parse tree. Second, to cope with sparse data, we smooth the lexical parameters according to their underlying word similarities using Laplacian Regularization. To demonstrate the benefits of our approach, we consider the problem of parsing Chinese treebank data using only lexical features, that is, without part-of-speech tags or grammatical categories. We achieve state of the art performance, improving upon current large margin approaches.

CONTEXT: Artificial intelligence (AI) is increasingly being recognized as having potential importance to primary care (PC). However, there is a gap in our understanding about where to focus efforts related to AI for PC settings, especially given the current COVID-19 pandemic. OBJECTIVE: To identify current priority areas for AI and PC in Ontario, Canada. STUDY DESIGN: Multi-stakeholder engagement event with facilitated small and large group discussions. A nominal group technique process was used to identify and rank challenges in PC that AI may be able to support. Mentimeter software was used to allow real-time, anonymous and independent ranking from all participants. A final list of priority areas for AI and PC, with key considerations, was derived based on ranked items and small group discussion notes. SETTING: Ontario, Canada. POPULATION STUDIED: Digital health and PC stakeholders. OUTCOME MEASURES: N/A. RESULTS: The event included 8 providers, 8 patient advisors, 4 decision makers, 3 digital health stakeholders, and 12 researchers. Nine priority areas for AI and PC were identified and ranked, which can be grouped into those intended to support physician (preventative care and risk profiling, clinical decision support, routine task support), patient (self-management of conditions, increased mental health care capacity and support), or system-level initiatives (administrative staff support, management and synthesis of information sources); and foundational areas that would support work on other priorities (improved communication between PC and AI stakeholders, data sharing and interoperability between providers). Small group discussions identified barriers and facilitators related to the priorities, including data availability, quality, and consent; legal and device certification issues; trust between people and technology; equity and the digital divide; patient centredness and usercentred design; and the need for funding to support collaborative research and pilot testing. Although identified areas do not explicitly mention COVID-19, participants were encouraged to think about what would be feasible and meaningful to accomplish within a few years, including considerations of the

Context: Electronic health records (EHR) provide an opportunity for developing decision support and other types of learning health system (LHS) initiatives. Careful understanding of the population of interest and how clients are represented in the data is essential for problem selection and for effective study design and analysis of data to solve the problem. The Alliance for Healthier Communities is one of the first primary care LHS in North America, serving complex, at-risk clients through Community Health Centres (CHCs) across Ontario, Canada. We propose that to properly understand their electronic health record data both simple statistical techniques commonly seen in descriptive epidemiology and more complex techniques from artificial intelligence will be useful.