Health Research and Policy


DATE: November 14, 2013
TIME: 1:15 - 3:00 pm
LOCATION: Medical School Office Building, Rm x303
TITLE: Reducing Mean Squared Error on the Estimation of Clonality and Abundances in Millions-Dimensional Data
Biomedical Informatics, Stanford

We focus on the use of ideas from point estimation to improve the estimation of simple statistics derived from extremely high (millions) dimensional data. While these questions have traditionally been primarily of relevance to ecology (with lower dimensions), modern experimental advances in molecular biology has opened the doors to analogous analyses in immunology, leading to biologically important findings.

Quantifying the clonal populations of B cells and T cells in individual lymphocyte repertoires is a topic of fundamental interest in immunology. The B cells and T cells of the adaptive immune system exhibit diverse and distinctive rearrangements of antigen receptor genes in their genomes, and undergo extensive clonal expansion in response to antigenic stimuli. Similar challenges of measurement arise in other areas of modern biology, such as estimating the contribution of different microbial species to microbiome populations, studying subclone populations of cells in a cancer, or evaluating the distribution of population members in synthetic libraries of macromolecules.

Modern DNA sequencing technology is a powerful means to study clonal features of complex molecular populations. Here, we address the problem of quantifying clonal contributions to B cell repertoires by deriving, implementing, and evaluating an estimator of a summary clonality score, using data from replicate sequencing libraries generated from a sample. In simulations, this approach reduces the mean squared error (MSE) of the clonality estimation by up to 99.7% compared to an estimator used in several recent studies. This approach also makes marked improvements in the estimation of the underlying clonal abundances.

Stanford Medicine Resources:

Footer Links: