Why do you think your paper is highly cited?

“Our paper was the first to address this debate from a statistical perspective. In particular, we outlined precise conditions under which pooling would be useful.”

The main contribution of the paper is that it provides something important which had been lacking from previous pooling debates, namely, a statistical framework defining precise conditions under which pooling is useful.

Microarrays are tools that allow for the measurement of mRNA abundance (or gene expression) levels in thousands of genes simultaneously. A disadvantage is cost—microarrays are expensive. Our paper addresses the utility of an experimental design that can decrease overall experimental cost.

A simple experiment comparing the gene expression profiles between two groups consists of gathering individuals in each of the groups and using a microarray to obtain each individual’s expression profile. Oftentimes, individual specific gene expression levels are not of interest, but rather average levels within a particular group. Those in favor of pooling note that when this is the case, assessing each individual’s expression level might not be necessary. Instead, they argue that average expression levels can be obtained if biological samples are first pooled and then the pool measured using a single microarray. When arrays are expensive relative to samples, this can greatly reduce cost. An argument against pooling is that aberrant samples might skew the average, making real differences harder to find. There is also concern that pooling could mask biological variation. Our paper was the first to address this debate from a statistical perspective. In particular, we outlined precise conditions under which pooling would be useful. We made no conjectures about whether or not the conditions hold for microarray data; we simply formalized the arguments on each side of the debate. Our more recent work, Kendziorski et al., PNAS, 102(12): 4252-4257, 2005, addresses the question of whether or not the conditions hold for experimental data.

I have had the privilege of working with many scientists here at UW-Madison and across the country. The questions I address, both theoretically and empirically, almost always arise from these collaborations. A few years ago, a basic question common to many of my collaborations was: "can we pool samples to save money?" Although the question is simply stated, the debate among my collaborators mimicked that in the literature. Some were strongly for; others strongly against. I became very interested in this problem since many scientists could not agree on the answer.