Carlo Berzuini

Following a moving target—Monte Carlo inference for dynamic Bayesian models

Markov chain Monte Carlo (MCMC) sampling is a numerically intensive simulation technique which has greatly improved the practicality of Bayesian inference and prediction. However, MCMC sampling is too slow to be of practical use in problems involving a large number of posterior (target) distributions, as in dynamic modelling and predictive model selection. Alternative simulation techniques for tracking moving target distributions, known as particle filters, which combine importance sampling, importance resampling and MCMC sampling, tend to suffer from a progressive degeneration as the target sequence evolves. We propose a new technique, based on these same simulation methodologies, which does not suffer from this progressive degeneration.

Genetic determinants of ulcerative colitis include the ECM1 locus and five loci implicated in Crohn's disease

We report results of a nonsynonymous SNP scan for ulcerative colitis and identify a previously unknown susceptibility locus at ECM1. We also show that several risk loci are common to ulcerative colitis and Crohns disease (IL23R, IL12B, HLA, NKX2-3 and MST1), whereas autophagy genes ATG16L1 and IRGM, along with NOD2 (also known as CARD15), are specific for Crohns disease. These data provide the first detailed illustration of the genetic relationship between these common inflammatory bowel diseases.

Dynamic conditional independence models and Markov chain Monte Carlo methods

Abstract In dynamic statistical modeling situations, observations arise sequentially, causing the model to expand by progressive incorporation of new data items and new unknown parameters. For example, in clinical monitoring, patients and data arrive sequentially, and new patient-specific parameters are introduced with each new patient. Markov chain Monte Carlo (MCMC) might be used for continuous updating of the evolving posterior distribution, but would need to be restarted from scratch at each expansion stage. Thus MCMC methods are often too slow for real-time inference in dynamic contexts. By combining MCMC with importance resampling, we show how real-time sequential updating of posterior distributions can be effected. The proposed dynamic sampling algorithms use posterior samples from previous updating stages and exploit conditional independence between groups of parameters to allow samples of parameters no longer of interest to be discarded, such as when a patient dies or is discharged. We apply the ...

No evidence of association between prothrombotic gene polymorphisms and the development of acute myocardial infarction at a young age

Background—We investigated the association between 9 polymorphisms of genes encoding hemostasis factors and myocardial infarction in a large sample of young patients chosen because they have less coronary atherosclerosis than older patients, and thus their disease is more likely to be related to a genetic predisposition to a prothrombotic state. Methods and Results—This nationwide case-control study involved 1210 patients who had survived a first myocardial infarction at an age of <45 years who underwent coronary arteriography in 125 coronary care units and 1210 healthy subjects matched for age, sex, and geographical origin. None of the 9 polymorphisms of genes encoding proteins involved in coagulation (G-455A &bgr;-fibrinogen: OR, 1.0; CI, 0.8 to 1.2; G1691A factor V: OR, 1.1; CI, 0.6 to 2.1; G20210A factor II: OR, 1.0; CI, 0.5 to 1.9; and G10976A factor VII: OR, 1.0; CI, 0.8 to 1.3), platelet function (C807T glycoprotein Ia: OR, 1.1; CI, 0.9 to 1.3; and C1565T glycoprotein IIIa: OR, 0.9; CI, 0.8 to 1.2), fibrinolysis (G185T factor XIII: OR, 1.2; CI, 0.9 to 1.6; and 4G/5G plasminogen activator inhibitor type 1: OR, 0.9; CI, 0.7 to 1.2), or homocysteine metabolism (C677T methylenetetrahydrofolate reductase: OR, 0.9; CI, 0.8 to 1.1) were associated with an increased or decreased risk of myocardial infarction. Conclusions—This study provides no evidence supporting an association between 9 polymorphisms of genes encoding proteins involved in hemostasis and the occurrence of premature myocardial infarction or protection against it.

A prognostic classification of myelofibrosis with myeloid metaplasia.

The dependence of survival time on a set of prognostic factors was explored by means of Coxs regression model in 137 cases of myelofibrosis with myeloid metaplasia (MMM). The following parameters recorded at diagnosis proved to be important independent indicators of a poor prognosis: a higher value for age, a lower value for Hb concentration, a higher value for immature myeloid cells in peripheral blood (IMC), a lower value for total erythroid iron turnover (TEIT), and a bone marrow red cell aplasia (RCA). A prognostic classification tree was constructed whose terminal nodes (risk groups), described by simple logical conditions upon important indicators, were characterized by significantly different expected survival. The two extreme risk groups lend themselves to a simple, but complete description. The low‐risk group (19.7% of the sample) comprises cases who had the diagnosis of MMM before age 45 and a number of IMC constantly lower than 24%. The actuarial proportion of patients surviving at 15 years was 100%. The high‐risk group (29.9% of cases) comprises patients with age greater than 45 and Hb lower than 13 g/dl, associated with RCA, or with a relatively decreased erythropoiesis (TEIT lower than 2 times the normal) or with IMC greater than 24%. Seven out of the 11 who died within this group developed blastic crisis. Median survival time of the group was 69 months.

Statistical analysis of domains in interacting protein pairs

MOTIVATION Several methods have recently been developed to analyse large-scale sets of physical interactions between proteins in terms of physical contacts between the constituent domains, often with a view to predicting new pairwise interactions. Our aim is to combine genomic interaction data, in which domain-domain contacts are not explicitly reported, with the domain-level structure of individual proteins, in order to learn about the structure of interacting protein pairs. Our approach is driven by the need to assess the evidence for physical contacts between domains in a statistically rigorous way. RESULTS We develop a statistical approach that assigns p-values to pairs of domain superfamilies, measuring the strength of evidence within a set of protein interactions that domains from these superfamilies form contacts. A set of p-values is calculated for SCOP superfamily pairs, based on a pooled data set of interactions from yeast. These p-values can be used to predict which domains come into contact in an interacting protein pair. This predictive scheme is tested against protein complexes in the Protein Quaternary Structure (PQS) database, and is used to predict domain-domain contacts within 705 interacting protein pairs taken from our pooled data set.

Predicting secondary progression in relapsing-remitting multiple sclerosis : a Bayesian analysis

With the aid of a Bayesian statistical model of the natural course of relapsing remitting Multiple Sclerosis (MS), we identify short-term clinical predictors of long-term evolution of the disease, with particular focus on predicting onset of secondary progressive course (failure event) on the basis of patient information available at an early stage of disease. The model specifies the full joint probability distribution for a set of variables including early indicator variables (observed during the early stage of disease), intermediate indicator variables (observed throughout the course of disease, prefailure) and the time to failure. Our model treats the intermediate indicators as a surrogate response event, so that in right-censored patients, these indicators provide supplementary information pointing towards the unobserved failure times. Moreover, the full probability modelling approach allows the considerable uncertainty which affects certain early indicators, such as the early relapse rates, to be incorporated in the analysis. With such a model, the ability of early indicators to predict failure can be assessed more accurately and reliably, and explained in terms of the relationship between early and intermediate indicators. Moreover, a model with the aforementioned features allows us to characterize the pattern of disease course in high-risk patients, and to identify short-term manifestations which are strongly related to long-term evolution of disease, as potential surrogate responses in clinical trials. Our analysis is based on longitudinal data from 186 MS patients with a relapsing-remitting initial course. The following important early predictors of the time to progression emerged: age; number of neurological functional systems (FSs) involved; sphincter, or motor, or motor-sensory symptoms; presence of sequelae after onset. During the first 3 years of follow up, to reach EDSS> or =4 outside relapse, to have sphincter or motor relapses and to reach moderate pyramidal involvement were also found to be unfavourable prognostic factors.

Interaction Between Exercise Training and Ejection Fraction in Predicting Prognosis After a First Myocardial Infarction

BACKGROUND Although recent meta-analysis trials have shown that exercise training may improve survival after myocardial infarction, the mechanism of this beneficial effect is still unknown. The purpose of this study was to detect possible interactions between exercise training and predictors of prognosis after a first myocardial infarction. METHODS AND RESULTS Patients with uneventful clinical courses after a first myocardial infarction were randomly assigned to a 4-week training period (125 patients, group 1) or to a control group (131 patients, group 2). Before randomization, all patients underwent a symptom-limited exercise test (28 +/- 2 days after myocardial infarction), 24-hour Holter monitoring, and coronary arteriography (31 +/- 3 days after the acute episode). After a mean follow-up period of 34.5 months, 18 patients had cardiac deaths (5 in group 1 and 13 in group 2). Multivariate analysis by Cox regression model showed that ejection fraction was the only independent prognostic indicator (P = .03). Evidence existed of an interaction between ejection fraction and exercise training, showing an effect of physical training on survival that depended on the patients ejection fraction. Among patients with ejection fractions < 41%, the relative risk for an untrained patient was 8.63 times higher than for a trained patient (P = .04), whereas for ejection fractions > 40%, the estimated risks for trained and untrained patients were similar. CONCLUSIONS These data show that exercise training may prolong survival in post-myocardial infarction patients with depressed left ventricular function. A randomized trial in such patients seems warranted.

Mapping the platelet profile for functional genomic studies and demonstration of the effect size of the GP6 locus

Summary.  Background: Evidence suggests the wide variation in platelet response within the population is genetically controlled. Unraveling the complex relationship between sequence variation and platelet phenotype requires accurate and reproducible measurement of platelet response. Objective: To develop a methodology suitable for measuring signaling pathway‐specific platelet phenotype, to use this to measure platelet response in a large cohort, and to demonstrate the effect size of sequence variation in a relevant model gene. Methods: Three established platelet assays were evaluated: mobilization of [Ca2+]i, aggregometry and flow cytometry, each in response to adenosine 5′‐diphosphate (ADP) or the glycoprotein (GP) VI‐specific crosslinked collagen‐related peptide (CRP). Flow cytometric measurement of fibrinogen binding and P‐selectin expression in response to a single, intermediate dose of each agonist gave the best combination of reproducibility and inter‐individual variability and was used to measure the platelet response in 506 healthy volunteers. Pathway specificity was ensured by blocking the main subsidiary signaling pathways. Results: Individuals were identified who were hypo‐ or hyper‐responders for both pathways, or who had differential responses to the two agonists, or between outcomes. 89 individuals, retested three months later using the same methodology, showed high concordance between the two visits in all four assays (r2 = 0.872, 0.868, 0.766 and 0.549); all subjects retaining their phenotype at recall. The effect of sequence variation at the GP6 locus accounted for ∼35% of the variation in the CRP‐XL response. Conclusion: Genotyping‐phenotype association studies in a well‐characterized, large cohort provides a powerful strategy to measure the effect of sequence variation in genes regulating the platelet response.

Contribution of TNFSF15 gene variants to Crohn's disease susceptibility confirmed in UK population

Background: Identification of Crohns disease (CD)‐associated genetic variants is key to understanding pathogenic pathways underlying disease susceptibility. Recent reports of an association between TNFSF15 variants and CD have been modestly replicated in European populations, suggesting heterogeneity at this locus with stronger CD association in Japanese than European populations. Methods: We investigated the association between variants in TNFSF15 and CD in 756 CD patients and 636 controls. Disease subphenotype associations were also investigated. Results: TNFSF15 single nucleotide polymorphism (SNP) variants were associated with CD in our panel with peak odds ratio (OR) 1.2 (95% confidence interval [CI] 1.01–1.41) P = 0.033. The presence of a risk haplotype was replicated for the first time in a European population (frequency 67% in cases and 61% in controls) OR = 1.44 (95% CI 1.23–1.68) P = 0.00012. This result mirrors the UK panel in the index study (Yamazaki et al [2005] Hum Mol Genet 14:3499–3506) but is less significant than that reported in Japanese populations. There was no evidence of association with any individual CD subphenotype. Conclusions: Variants in TNFSF15 contribute to overall CD susceptibility in European populations, although to a lesser extent than that seen in the Japanese. Further studies to define the precise disease‐causing variants as well as targeted functional studies are now required in human CD as TNFSF15 is a potential target for biological therapies.