Democratizing data at Novartis through clinical trial data access.

Enabling broad access and usage of clinical trial data within biopharmaceutical companies has historically been impeded by technical, cultural, and policy hurdles. Novartis has attempted to address this comprehensively through a program called data42; here, we explore how a diverse set of enterprise-wide stakeholders formulated a risk-based data access approach to streamline access to anonymized clinical trial data and vastly improved its use by authorized research and development (R&D) associates within the company. The result is that most Novartis clinical trial data requests, from internal associates, can now be automatically approved. The process of developing this framework and its impact on Novartis and the broader industry are explored and discussed.

Translational precision medicine: an industry perspective.

In the era of precision medicine, digital technologies and artificial intelligence, drug discovery and development face unprecedented opportunities for product and business model innovation, fundamentally changing the traditional approach of how drugs are discovered, developed and marketed. Critical to this transformation is the adoption of new technologies in the drug development process, catalyzing the transition from serendipity-driven to data-driven medicine. This paradigm shift comes with a need for both translation and precision, leading to a modern Translational Precision Medicine approach to drug discovery and development. Key components of Translational Precision Medicine are multi-omics profiling, digital biomarkers, model-based data integration, artificial intelligence, biomarker-guided trial designs and patient-centric companion diagnostics. In this review, we summarize and critically discuss the potential and challenges of Translational Precision Medicine from a cross-industry perspective.

Artificial intelligence and machine learning in clinical development: a translational perspective.

Future of clinical development is on the verge of a major transformation due to convergence of large new digital data sources, computing power to identify clinically meaningful patterns in the data using efficient artificial intelligence and machine-learning algorithms, and regulators embracing this change through new collaborations. This perspective summarizes insights, recent developments, and recommendations for infusing actionable computational evidence into clinical development and health care from academy, biotechnology industry, nonprofit foundations, regulators, and technology corporations. Analysis and learning from publically available biomedical and clinical trial data sets, real-world evidence from sensors, and health records by machine-learning architectures are discussed. Strategies for modernizing the clinical development process by integration of AI- and ML-based digital methods and secure computing technologies through recently announced regulatory pathways at the United States Food and Drug Administration are outlined. We conclude by discussing applications and impact of digital algorithmic evidence to improve medical care for patients.

Evaluating the contribution of rare variants to type 2 diabetes and related traits using pedigrees.

A major challenge in evaluating the contribution of rare variants to complex disease is identifying enough copies of the rare alleles to permit informative statistical analysis. To investigate the contribution of rare variants to the risk of type 2 diabetes (T2D) and related traits, we performed deep whole-genome analysis of 1,034 members of 20 large Mexican-American families with high prevalence of T2D. If rare variants of large effect accounted for much of the diabetes risk in these families, our experiment was powered to detect association. Using gene expression data on 21,677 transcripts for 643 pedigree members, we identified evidence for large-effect rare-variant cis-expression quantitative trait loci that could not be detected in population studies, validating our approach. However, we did not identify any rare variants of large effect associated with T2D, or the related traits of fasting glucose and insulin, suggesting that large-effect rare variants account for only a modest fraction of the genetic risk of these traits in this sample of families. Reliable identification of large-effect rare variants will require larger samples of extended pedigrees or different study designs that further enrich for such variants.

A Bayesian mathematical model of motor and cognitive outcomes in Parkinson's disease.

BACKGROUND: There are few established predictors of the clinical course of PD. Prognostic markers would be useful for clinical care and research. OBJECTIVE: To identify predictors of long-term motor and cognitive outcomes and rate of progression in PD. METHODS: Newly diagnosed PD participants were followed for 7 years in a prospective study, conducted at 55 centers in the United States and Canada. Analyses were conducted in 244 participants with complete demographic, clinical, genetic, and dopamine transporter imaging data. Machine learning dynamic Bayesian graphical models were used to identify and simulate predictors and outcomes. The outcomes rate of cognition changes are assessed by the Montreal Cognitive Assessment scores, and rate of motor changes are assessed by UPDRS part-III. RESULTS: The most robust and consistent longitudinal predictors of cognitive function included older age, baseline Unified Parkinson's Disease Rating Scale (UPDRS) parts I and II, Schwab and England activities of daily living scale, striatal dopamine transporter binding, and SNP rs11724635 in the gene BST1. The most consistent predictor of UPDRS part III was baseline level of activities of daily living (part II). Key findings were replicated using long-term data from an independent cohort study. CONCLUSIONS: Baseline function near the time of Parkinson's disease diagnosis, as measured by activities of daily living, is a consistent predictor of long-term motor and cognitive outcomes. Additional predictors identified may further characterize the expected course of Parkinson's disease and suggest mechanisms underlying disease progression. The prognostic model developed in this study can be used to simulate the effects of the prognostic variables on motor and cognitive outcomes, and can be replicated and refined with data from independent longitudinal studies.

Reverse Engineering and Evaluation of Prediction Models for Progression to Type 2 Diabetes: An Application of Machine Learning Using Electronic Health Records.

BACKGROUND: Application of novel machine learning approaches to electronic health record (EHR) data could provide valuable insights into disease processes. We utilized this approach to build predictive models for progression to prediabetes and type 2 diabetes (T2D). METHODS: Using a novel analytical platform (Reverse Engineering and Forward Simulation [REFS]), we built prediction model ensembles for progression to prediabetes or T2D from an aggregated EHR data sample. REFS relies on a Bayesian scoring algorithm to explore a wide model space, and outputs a distribution of risk estimates from an ensemble of prediction models. We retrospectively followed 24 331 adults for transitions to prediabetes or T2D, 2007-2012. Accuracy of prediction models was assessed using an area under the curve (AUC) statistic, and validated in an independent data set. RESULTS: Our primary ensemble of models accurately predicted progression to T2D (AUC = 0.76), and was validated out of sample (AUC = 0.78). Models of progression to T2D consisted primarily of established risk factors (blood glucose, blood pressure, triglycerides, hypertension, lipid disorders, socioeconomic factors), whereas models of progression to prediabetes included novel factors (high-density lipoprotein, alanine aminotransferase, C-reactive protein, body temperature; AUC = 0.70). CONCLUSIONS: We constructed accurate prediction models from EHR data using a hypothesis-free machine learning approach. Identification of established risk factors for T2D serves as proof of concept for this analytical approach, while novel factors selected by REFS represent emerging areas of T2D research. This methodology has potentially valuable downstream applications to personalized medicine and clinical research.

Whole-genome sequencing of asian lung cancers: second-hand smoke unlikely to be responsible for higher incidence of lung cancer among Asian never-smokers.

Asian nonsmoking populations have a higher incidence of lung cancer compared with their European counterparts. There is a long-standing hypothesis that the increase of lung cancer in Asian never-smokers is due to environmental factors such as second-hand smoke. We analyzed whole-genome sequencing of 30 Asian lung cancers. Unsupervised clustering of mutational signatures separated the patients into two categories of either all the never-smokers or all the smokers or ex-smokers. In addition, nearly one third of the ex-smokers and smokers classified with the never-smoker-like cluster. The somatic variant profiles of Asian lung cancers were similar to that of European origin with G.C>T.A being predominant in smokers. We found EGFR and TP53 to be the most frequently mutated genes with mutations in 50% and 27% of individuals, respectively. Among the 16 never-smokers, 69% had an EGFR mutation compared with 29% of 14 smokers/ex-smokers. Asian never-smokers had lung cancer signatures distinct from the smoker signature and their mutation profiles were similar to European never-smokers. The profiles of Asian and European smokers are also similar. Taken together, these results suggested that the same mutational mechanisms underlie the etiology for both ethnic groups. Thus, the high incidence of lung cancer in Asian never-smokers seems unlikely to be due to second-hand smoke or other carcinogens that cause oxidative DNA damage, implying that routine EGFR testing is warranted in the Asian population regardless of smoking status.

MYD88 L265P somatic mutation in Waldenström's macroglobulinemia.

BACKGROUND: Waldenström's macroglobulinemia is an incurable, IgM-secreting lymphoplasmacytic lymphoma (LPL). The underlying mutation in this disorder has not been delineated. METHODS: We performed whole-genome sequencing of bone marrow LPL cells in 30 patients with Waldenström's macroglobulinemia, with paired normal-tissue and tumor-tissue sequencing in 10 patients. Sanger sequencing was used to validate the findings in samples from an expanded cohort of patients with LPL, those with other B-cell disorders that have some of the same features as LPL, and healthy donors. RESULTS: Among the patients with Waldenström's macroglobulinemia, a somatic variant (T→C) in LPL cells was identified at position 38182641 at 3p22.2 in the samples from all 10 patients with paired tissue samples and in 17 of 20 samples from patients with unpaired samples. This variant predicted an amino acid change (L265P) in MYD88, a mutation that triggers IRAK-mediated NF-κB signaling. Sanger sequencing identified MYD88 L265P in tumor samples from 49 of 54 patients with Waldenström's macroglobulinemia and in 3 of 3 patients with non-IgM-secreting LPL (91% of all patients with LPL). MYD88 L265P was absent in paired normal tissue samples from patients with Waldenström's macroglobulinemia or non-IgM LPL and in B cells from healthy donors and was absent or rarely expressed in samples from patients with multiple myeloma, marginal-zone lymphoma, or IgM monoclonal gammopathy of unknown significance. Inhibition of MYD88 signaling reduced IκBα and NF-κB p65 phosphorylation, as well as NF-κB nuclear staining, in Waldenström's macroglobulinemia cells expressing MYD88 L265P. Somatic variants in ARID1A in 5 of 30 patients (17%), leading to a premature stop or frameshift, were also identified and were associated with an increased disease burden. In addition, 2 of 3 patients with Waldenström's macroglobulinemia who had wild-type MYD88 had somatic variants in MLL2. CONCLUSIONS: MYD88 L265P is a commonly recurring mutation in patients with Waldenström's macroglobulinemia that can be useful in differentiating Waldenström's macroglobulinemia and non-IgM LPL from B-cell disorders that have some of the same features. (Funded by the Peter and Helen Bing Foundation and others.).

Relation of increased prebeta-1 high-density lipoprotein levels to risk of coronary heart disease.

Preß-1 high-density lipoprotein (HDL) plays a key role in reverse cholesterol transport by promoting cholesterol efflux. Our aims were (1) to test previous associations between preß-1 HDL and coronary heart disease (CHD) and (2) to investigate whether preß-1 HDL levels also are associated with risk of myocardial infarction (MI). Plasma preß-1 HDL was measured by an ultrafiltration-isotope dilution technique in 1,255 subjects recruited from the University of California-San Francisco Lipid and Cardiovascular Clinics and collaborating cardiologists. Preß-1 HDL was significantly and positively associated with CHD and MI even after adjustment for established risk factors. Inclusion of preß-1 HDL in a multivariable model for CHD led to a modest improvement in reclassification of subjects (net reclassification index 0.15, p = 0.01; integrated discrimination improvement 0.003, p = 0.2). In contrast, incorporation of preß-1 HDL into a risk model of MI alone significantly improved reclassification of subjects (net reclassification index 0.21, p = 0.008; integrated discrimination improvement 0.01, p = 0.02), suggesting that preß-1 HDL has more discriminatory power for MI than for CHD in our study population. In conclusion, these results confirm previous associations between preß-1 HDL and CHD in a large well-characterized clinical cohort. Also, this is the first study in which preß-1 HDL was identified as a novel and independent predictor of MI above and beyond traditional CHD risk factors.

Power in the phenotypic extremes: a simulation study of power in discovery and replication of rare variants.

Next-generation sequencing technologies are making it possible to study the role of rare variants in human disease. Many studies balance statistical power with cost-effectiveness by (a) sampling from phenotypic extremes and (b) utilizing a two-stage design. Two-stage designs include a broad-based discovery phase and selection of a subset of potential causal genes/variants to be further examined in independent samples. We evaluate three parameters: first, the gain in statistical power due to extreme sampling to discover causal variants; second, the informativeness of initial (Phase I) association statistics to select genes/variants for follow-up; third, the impact of extreme and random sampling in (Phase 2) replication. We present a quantitative method to select individuals from the phenotypic extremes of a binary trait, and simulate disease association studies under a variety of sample sizes and sampling schemes. First, we find that while studies sampling from extremes have excellent power to discover rare variants, they have limited power to associate them to phenotype­suggesting high false-negative rates for upcoming studies. Second, consistent with previous studies, we find that the effect sizes estimated in these studies are expected to be systematically larger compared with the overall population effect size; in a well-cited lipids study, we estimate the reported effect to be twofold larger. Third, replication studies require large samples from the general population to have sufficient power; extreme sampling could reduce the required sample size as much as fourfold. Our observations offer practical guidance for the design and interpretation of studies that utilize extreme sampling.

Multiple genes influence BMI on chromosome 7q31-34: the NHLBI Family Heart Study.

The National Heart, Lung, and Blood Institute Family Heart Study (FHS) genome-wide linkage scan identified a region of chromosome 7q31-34 with a lod score of 4.9 for BMI at D7S1804 (131.9 Mb). We report the results of linkage and association to BMI in this region for two independent FHS samples. The first sample includes 225 FHS pedigrees with evidence of linkage to 7q31-34, using 1,132 single-nucleotide polymorphisms (SNPs) and 7 microsatellites. The second represents a case-control sample (318 cases; BMI >25 and 325 controls; BMI <25) derived from unrelated FHS participants who were not part of the genome scan. The latter set was genotyped for 606 SNPs, including 37 SNPs with prior evidence for association in the linked families. Although variance components linkage analysis using only SNPs generated a peak lod score that coincided with the original linkage scan at 131.9 Mb, a conditional linkage analysis showed evidence of a second quantitative trait locus (QTL) near 143 cM influencing BMI. Three SNPs (rs161339, rs12673281, and rs1993068) located near the three genes pleiotrophin (PTN), diacylglycerol (DAG) kinase iota (DGK iota), and cholinergic receptor, muscarinic 2 (CHRM2) demonstrated significant association in both linked families (P = 0.0005, 0.002, and 0.03, respectively) and the case-control sample (P = 0.01, 0.0003, and 0.03, respectively), regardless of the genetic model tested. These findings suggest that several genes may be associated with BMI in the 7q31-34 region.

Leptin is associated with blood pressure and hypertension in women from the National Heart, Lung, and Blood Institute Family Heart Study.

Leptin is a key neuroendocrine hormone regulating food intake, metabolism, and fat accumulation, and it may also affect blood pressure and contribute to hypertension through sympathetic activation in the vasculature or at the renal level. Although previous studies have shown that the distribution of leptin is significantly different between males and females, as is the risk of hypertension between males and females, results regarding the role of leptin in the gender-specific regulation of blood pressure are controversial. Thus, we performed family-based association analyses in the National Heart, Lung, and Blood Institute Family Heart Study to test the hypothesis that LEPTIN gene variants and the plasma leptin level influence variability in blood pressure and the risk of hypertension differently by gender. We identified significant associations between LEPTIN single nucleotide polymorphisms with blood pressure and hypertension, but in postmenopausal women only. We also identified significant associations between plasma leptin levels and both blood pressure and hypertension in women. The current study supports a role for LEPTIN and plasma leptin levels in blood pressure regulation in women. It also provides insight into the gender differences in hypertension, as well as the differential distribution and activity of leptin in men and women.

The Gly2019Ser mutation in LRRK2 is not fully penetrant in familial Parkinson's disease: the GenePD study.

BACKGROUND: We report age-dependent penetrance estimates for leucine-rich repeat kinase 2 (LRRK2)-related Parkinson's disease (PD) in a large sample of familial PD. The most frequently seen LRRK2 mutation, Gly2019Ser (G2019S), is associated with approximately 5 to 6% of familial PD cases and 1 to 2% of idiopathic cases, making it the most common known genetic cause of PD. Studies of the penetrance of LRRK2 mutations have produced a wide range of estimates, possibly due to differences in study design and recruitment, including in particular differences between samples of familial PD versus sporadic PD. METHODS: A sample, including 903 affected and 58 unaffected members from 509 families ascertained for having two or more PD-affected members, 126 randomly ascertained PD patients and 197 controls, was screened for five different LRRK2 mutations. Penetrance was estimated in families of LRRK2 carriers with consideration of the inherent bias towards increased penetrance in a familial sample. RESULTS: Thirty-one out of 509 families with multiple cases of PD (6.1%) were found to have 58 LRRK2 mutation carriers (6.4%). Twenty-nine of the 31 families had G2019S mutations while two had R1441C mutations. No mutations were identified among controls or unaffected relatives of PD cases. Nine PD-affected relatives of G2019S carriers did not carry the LRRK2 mutation themselves. At the maximum observed age range of 90 to 94 years, the unbiased estimated penetrance was 67% for G2019S families, compared with a baseline PD risk of 17% seen in the non-LRRK2-related PD families. CONCLUSION: Lifetime penetrance of LRRK2 estimated in the unascertained relatives of multiplex PD families is greater than that reported in studies of sporadically ascertained LRRK2 cases, suggesting that inherited susceptibility factors may modify the penetrance of LRRK2 mutations. In addition, the presence of nine PD phenocopies in the LRRK2 families suggests that these susceptibility factors may also increase the risk of non-LRRK2-related PD. No differences in penetrance were found between men and women, suggesting that the factors that influence penetrance for LRRK2 carriers are independent of the factors which increase PD prevalence in men.

Polymorphisms near EXOC4 and LRGUK on chromosome 7q32 are associated with Type 2 Diabetes and fasting glucose; the NHLBI Family Heart Study.

BACKGROUND: The chromosome 7q32 region is linked to metabolic syndrome and obesity related traits in the Family Heart Study. As part of a fine mapping study of the region, we evaluated the relationship of polymorphisms to fasting glucose levels and Type 2 diabetes. METHODS: Thirty-nine HapMap defined tag SNPs in a 1.08 Mb region and a novel deletion polymorphism were genotyped in 2,603 participants of the NHLBI Family Heart Study (FHS). Regression modeling, adjusting for BMI, age, sex, smoking and the TCF7L2 polymorphism, was used to evaluate the association of these polymorphisms with T2D and fasting glucoses levels. RESULTS: The deletion polymorphism confers a protective effect for T2D, with homozygous deletion carriers having a 53% reduced risk compared to non-deleted carriers. Among non-diabetics, the deletion was significantly associated with lower fasting glucose levels in men (p = 0.038) but not women (p = 0.118). In addition, seven SNPs near the deletion were significantly associated (p < 0.01) to diabetes. CONCLUSION: Chromosome 7q32 contains both SNPs and a deletion that were associated to T2D. Although the deletion region contains several islands of strongly conserved sequence, it is not known to contain a transcribed gene. The closest nearby gene, EXOC4, is involved in insulin-stimulated glucose transport and may be a candidate for this association. Further work is needed to determine if the deletion represents a functional variant or may be in linkage disequilibrium with a functional mutation influencing EXOC4 or another nearby gene.

Transcriptional profiles of human epithelial cells in response to heat: computational evidence for novel heat shock proteins.

We hypothesized that broad-scale expression profiling would provide insight into the regulatory pathways that control gene expression in response to stress and potentially identify novel heat-responsive genes. HEp2 cells, a human malignant epithelial cell line, were heated at 37 degrees C to 43 degrees C for 60 min to gauge the heat shock response, using as a proxy inducible Hsp70 quantified by Western blot analysis. Based on these results, microarray experiments were conducted at 37 degrees C, 40 degrees C, 41 degrees C, 42 degrees C, and 43 degrees C. Using linear modeling, we compared the sets of microarrays at 40 degrees C, 41 degrees C, 42 degrees C, and 43 degrees C with the 37 degrees C baseline temperature and took the union of the genes exhibiting differential gene expression signal to create two sets of "heat shock response" genes, each set reflecting either increased or decreased RNA abundance. Leveraging human and mouse orthologous alignments, we used the two lists of coexpressed genes to predict transcription factor binding sites in silico, including those for heat shock factor (HSF) 1 and HSF2 transcription factors. We discovered HSF1 and HSF2 binding sites in 15 genes not previously associated with the heat shock response. We conclude that microarray experiments coupled with upstream promoter analysis can be used to identify novel genes that respond to heat shock. Additional experiments are required to validate these putative heat shock proteins and facilitate a deeper understanding of the mechanisms involved during the stress response.

The Framingham Heart Study 100K SNP genome-wide association study resource: overview of 17 phenotype working group reports.

BACKGROUND: The Framingham Heart Study (FHS), founded in 1948 to examine the epidemiology of cardiovascular disease, is among the most comprehensively characterized multi-generational studies in the world. Many collected phenotypes have substantial genetic contributors; yet most genetic determinants remain to be identified. Using single nucleotide polymorphisms (SNPs) from a 100K genome-wide scan, we examine the associations of common polymorphisms with phenotypic variation in this community-based cohort and provide a full-disclosure, web-based resource of results for future replication studies. METHODS: Adult participants (n = 1345) of the largest 310 pedigrees in the FHS, many biologically related, were genotyped with the 100K Affymetrix GeneChip. These genotypes were used to assess their contribution to 987 phenotypes collected in FHS over 56 years of follow up, including: cardiovascular risk factors and biomarkers; subclinical and clinical cardiovascular disease; cancer and longevity traits; and traits in pulmonary, sleep, neurology, renal, and bone domains. We conducted genome-wide variance components linkage and population-based and family-based association tests. RESULTS: The participants were white of European descent and from the FHS Original and Offspring Cohorts (examination 1 Offspring mean age 32 +/- 9 years, 54% women). This overview summarizes the methods, selected findings and limitations of the results presented in the accompanying series of 17 manuscripts. The presented association results are based on 70,897 autosomal SNPs meeting the following criteria: minor allele frequency > or + 10%, genotype call rate > or = 80%, Hardy-Weinberg equilibrium p-value > or = 0.001, and satisfying Mendelian consistency. Linkage analyses are based on 11,200 SNPs and short-tandem repeats. Results of phenotype-genotype linkages and associations for all autosomal SNPs are posted on the NCBI dbGaP website at http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?id=phs000007 webcite. CONCLUSION: We have created a full-disclosure resource of results, posted on the dbGaP website, from a genome-wide association study in the FHS. Because we used three analytical approaches to examine the association and linkage of 987 phenotypes with thousands of SNPs, our results must be considered hypothesis-generating and need to be replicated. Results from the FHS 100K project with NCBI web posting provides a resource for investigators to identify high priority findings for replication.

Framingham Heart Study genome-wide association: results for pulmonary function measures.

BACKGROUND: Pulmonary function measures obtained by spirometry are used to diagnose chronic obstructive pulmonary disease (COPD) and are highly heritable. We conducted genome-wide association (GWA) analyses (Affymetrix 100K SNP GeneChip) for measures of lung function in the Framingham Heart Study. METHODS: Ten spirometry phenotypes including percent of predicted measures, mean spirometry measures over two examinations, and rates of change based on forced expiratory volume in one second (FEV1), forced vital capacity (FVC), forced expiratory flow from the 25th to 75th percentile (FEF25-75), the FEV1/FVC ratio, and the FEF25-75/FVC ratio were examined. Percent predicted phenotypes were created using each participant's latest exam with spirometry. Predicted lung function was estimated using models defined in the set of healthy never-smokers, and standardized residuals of percent predicted measures were created adjusting for smoking status, pack-years, and body mass index (BMI). All modeling was performed stratified by sex and cohort. Mean spirometry phenotypes were created using data from two examinations and adjusting for age, BMI, height, smoking and pack-years. Change in pulmonary function over time was studied using two to four examinations with spirometry to calculate slopes, which were then adjusted for age, height, smoking and pack-years. RESULTS: Analyses were restricted to 70,987 autosomal SNPs with minor allele frequency > or = 10%, genotype call rate > or = 80%, and Hardy-Weinberg equilibrium p-value > or = 0.001. A SNP in the interleukin 6 receptor (IL6R) on chromosome 1 was among the best results for percent predicted FEF25-75. A non-synonymous coding SNP in glutathione S-transferase omega 2 (GSTO2) on chromosome 10 had top-ranked results studying the mean FEV1 and FVC measurements from two examinations. SNPs nearby the SOD3 and vitamin D binding protein genes, candidate genes for COPD, exhibited association to percent predicted phenotypes. CONCLUSION: GSTO2 and IL6R are credible candidate genes for association to pulmonary function identified by GWA. These and other observed associations warrant replication studies. This resource of GWA results for pulmonary function measures is publicly available at http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?id=phs000007 webcite.