Details
- [Return to Previous Page]
- Student Seminar - Hojjat Salmasian and Alexander Lachman
- Date : February 28th, 2013
Time : 4:00PM - 5:00 PM
Location : VC-5
Event Speaker : Hojjat Salmasian and Alexander Lachman -
Hojjat Salmasian
Title: Developing automated methods for identifying and reducing inappropriate medical care
Abstract: Inappropriate medical care is an urgent health care challenge, with enormous impact on costs and patient safety. At his been estimated that approximately 30% of the total health care expenditure. Numerous articles have reported on the pervasiveness of inappropriate care and on noncompliance concerning established protocols for procedures. In this presentation, I will provide a concise review of the problem and proposed strategies, and describe my research plans to develop an automated solution, as week as the challenges involved.
Alexander Lachman
Title: Predicting genome-wide expression profiles using regularized regression and predicting drug mode of action
Abstract: In recent years there is increasing interest in computationally identifying the mode of action (MOA) of drugs from their measured effect on gene expression. To this end large scale studies have been performed to measure the effect of single drug perturbations on a variety of cell types. One such publicly available data resource is provided by the connectivity map (CMAP), which uncovers similarities between drugs and infers their MOA. However, given thousands of chemical compounds in combination with a diverse set of cell lines to be studied, the number of experiments that have to be performed extends beyond the resources available today. A novel approach to reduce the cost of creating such datasets is by using Luminex bead assays. This technology offers a compromise between sample size and genes covered, by allowing to measure 1000 genes simultaneously at very low cost. Restricting the view to 1000 genes poses a limiting perspective of the mechanisms underlying cellular responses.
Our goal is to reliably approximate the remaining gene expression profile from a large dataset that is currently in development as part of the Large Scale Expression Analysis of Cellular States (LINCS) project. To achieve this we use elastic-net to build a linear model for each gene not measured by the Luminex bead assay.
