Tuesday, April 26, 2016 - 11:00
Derek Chiu & Xiaoting Ding, UBC Statistics MSc. candidates
Room 4192, Earth Sciences Building, 2207 Main Mall
Two 30-minute talks will be given to discuss the co-op experience.
Talk 1 (11:00am - 11:30am)
Speaker: Derek Chiu, UBC Statistics Master's Student (Co-op)
Title: Meta-consensus clustering algorithm for HGSC subtype discovery
Abstract: High grade serous carcinoma (HGSC) is the most common form of ovarian cancer, and can be divided into several subtypes, each with distinct pathological properties. Cluster analysis can be used to group patient samples based on similarity in gene expression data. The goal is to classify patients into these subtypes for more targeted treatment. However, there are two sources of variability we need to take into account. First, there is variability between different runs of a clustering algorithm. Also, different clustering algorithms produce different results. We constructed a “meta-consensus” clustering method adapted from Monti et al. to handle some of these limitations. The approach pools together results from different clustering methods and replications before arriving at a final cluster assignment. We also used internal clustering indices to assess performance on the notions of stability, compactness, and separability. This talk describes analyses performed at OVCARE, BC Cancer Agency as part of my Co-op work term.
Talk 2 (11:30am - 12:00pm)
Speaker: Xiaoting Ding, UBC Statistics Master's Student (Co-op)
Title: The effect of different case definition of current smoking on the discovery of smoking -related blood gene expression signatures in COPD.
Abstract: Smokers and people exposed to secondhand smoke are at most important risk for chronic obstructive pulmonary disease (COPD). Some findings showed that high proportions of smoking patients with lung disease underestimates or denies smoking. Our goal was to find differences in blood gene expression between current smokers and former smokers, where smoking status was defined by self-reported status, objective measurement (exhaled carbon monoxide) and the combination of these two. Our hypothesis was to use the combination of self-reported and objective measurement to represent the smoking status was a better way to classify the phenotype, so that we could identify differential expressions more effectively.
Talk 1 (11:00am - 11:30am)
Speaker: Derek Chiu, UBC Statistics Master's Student (Co-op)
Title: Meta-consensus clustering algorithm for HGSC subtype discovery
Abstract: High grade serous carcinoma (HGSC) is the most common form of ovarian cancer, and can be divided into several subtypes, each with distinct pathological properties. Cluster analysis can be used to group patient samples based on similarity in gene expression data. The goal is to classify patients into these subtypes for more targeted treatment. However, there are two sources of variability we need to take into account. First, there is variability between different runs of a clustering algorithm. Also, different clustering algorithms produce different results. We constructed a “meta-consensus” clustering method adapted from Monti et al. to handle some of these limitations. The approach pools together results from different clustering methods and replications before arriving at a final cluster assignment. We also used internal clustering indices to assess performance on the notions of stability, compactness, and separability. This talk describes analyses performed at OVCARE, BC Cancer Agency as part of my Co-op work term.
Talk 2 (11:30am - 12:00pm)
Speaker: Xiaoting Ding, UBC Statistics Master's Student (Co-op)
Title: The effect of different case definition of current smoking on the discovery of smoking -related blood gene expression signatures in COPD.
Abstract: Smokers and people exposed to secondhand smoke are at most important risk for chronic obstructive pulmonary disease (COPD). Some findings showed that high proportions of smoking patients with lung disease underestimates or denies smoking. Our goal was to find differences in blood gene expression between current smokers and former smokers, where smoking status was defined by self-reported status, objective measurement (exhaled carbon monoxide) and the combination of these two. Our hypothesis was to use the combination of self-reported and objective measurement to represent the smoking status was a better way to classify the phenotype, so that we could identify differential expressions more effectively.