Adam Kowalczyk

Learning from structured data is becoming increasingly important. However, most prior work on kernel methods has focused on learning from attribute-value data. Only recently, research started investigating kernels for structured data. This paper considers kernels for multi-instance problems -- a class of concepts on individuals represented by sets. The main result of this paper is a kernel on multi-instance data that can be shown to separate positive and negative sets under natural assumptions. This kernel compares favorably with state of the art multi-instance learning algorithms in an empirical study. Finally, we give some concluding remarks and propose future work that might further improve the results.

We have performed plasma lipid profiling using liquid chromatography electrospray ionization tandem mass spectrometry on a population cohort of more than 1,000 individuals. From 10 μl of plasma we were able to acquire comparative measures of 312 lipids across 23 lipid classes and subclasses including sphingolipids, phospholipids, glycerolipids, and cholesterol esters (CEs) in 20 min. Using linear and logistic regression, we identified statistically significant associations of lipid classes, subclasses, and individual lipid species with anthropometric and physiological measures. In addition to the expected associations of CEs and triacylglycerol with age, sex, and body mass index (BMI), ceramide was significantly higher in males and was independently associated with age and BMI. Associations were also observed for sphingomyelin with age but this lipid subclass was lower in males. Lysophospholipids were associated with age and higher in males, but showed a strong negative association with BMI. Many of these lipids have previously been associated with chronic diseases including cardiovascular disease and may mediate the interactions of age, sex, and obesity with disease risk.

The relationship between lipid metabolism with prediabetes (impaired fasting glucose and impaired glucose tolerance) and type 2 diabetes mellitus is poorly defined. We hypothesized that a lipidomic analysis of plasma lipids might improve the understanding of this relationship. We performed lipidomic analysis measuring 259 individual lipid species, including sphingolipids, phospholipids, glycerolipids and cholesterol esters, on fasting plasma from 117 type 2 diabetes, 64 prediabetes and 170 normal glucose tolerant participants in the Australian Diabetes, Obesity and Lifestyle Study (AusDiab) then validated our findings on 1076 individuals from the San Antonio Family Heart Study (SAFHS). Logistic regression analysis of identified associations with type 2 diabetes (135 lipids) and prediabetes (134 lipids), after adjusting for multiple covariates. In addition to the expected associations with diacylglycerol, triacylglycerol and cholesterol esters, type 2 diabetes and prediabetes were positively associated with ceramide, and its precursor dihydroceramide, along with phosphatidylethanolamine, phosphatidylglycerol and phosphatidylinositol. Significant negative associations were observed with the ether-linked phospholipids alkylphosphatidylcholine and alkenylphosphatidylcholine. Most of the significant associations in the AusDiab cohort (90%) were subsequently validated in the SAFHS cohort. The aberration of the plasma lipidome associated with type 2 diabetes is clearly present in prediabetes, prior to the onset of type 2 diabetes. Lipid classes and species associated with type 2 diabetes provide support for a number of existing paradigms of dyslipidemia and suggest new avenues of investigation.

OBJECTIVE: Traditional risk factors for coronary artery disease (CAD) fail to adequately distinguish patients who have atherosclerotic plaques susceptible to instability from those who have more benign forms. Using plasma lipid profiling, this study aimed to provide insight into disease pathogenesis and evaluate the potential of lipid profiles to assess risk of future plaque instability. METHODS AND RESULTS: Plasma lipid profiles containing 305 lipids were measured on 220 individuals (matched healthy controls, n=80; stable angina, n=60; unstable coronary syndrome, n=80) using electrospray-ionisation tandem mass spectrometry. ReliefF feature selection coupled with an L2-regularized logistic regression based classifier was used to create multivariate classification models which were verified via 3-fold cross-validation (1000 repeats). Models incorporating both lipids and traditional risk factors provided improved classification of unstable CAD from stable CAD (C-statistic=0.875, 95% CI 0.874-0.877) compared with models containing only traditional risk factors (C-statistic=0.796, 95% CI 0.795-0.798). Many of the lipids identified as discriminatory for unstable CAD displayed an association with disease acuity (severity), suggesting that they are antecedents to the onset of acute coronary syndrome. CONCLUSION: Plasma lipid profiling may contribute to a new approach to risk stratification for unstable CAD.

Emerging evidence suggests that poor glycemic control mediates post-translational modifications to the H3 histone tail. We are only beginning to understand the dynamic role of some of the diverse epigenetic changes mediated by hyperglycemia at single loci, yet elevated glucose levels are thought to regulate genome-wide changes, and this still remains poorly understood. In this article we describe genome-wide histone H3K9/K14 hyperacetylation and DNA methylation maps conferred by hyperglycemia in primary human vascular cells. Chromatin immunoprecipitation (ChIP) as well as CpG methylation (CpG) assays, followed by massive parallel sequencing (ChIP-seq and CpG-seq) identified unique hyperacetylation and CpG methylation signatures with proximal and distal patterns of regionalization associative with gene expression. Ingenuity knowledge-based pathway and gene ontology analyses indicate that hyperglycemia significantly affects human vascular chromatin with the transcriptional up-regulation of genes involved in metabolic and cardiovascular disease. We have generated the first installment of a reference collection of hyperglycemia-induced chromatin modifications using robust and reproducible platforms that allow parallel sequencing-by-synthesis of immunopurified content. We uncover that hyperglycemia-mediated induction of genes and pathways associated with endothelial dysfunction occur through modulation of acetylated H3K9/K14 inversely correlated with methyl-CpG content.