\magnification=1200
\baselineskip=20pt
\nopagenumbers
\font\big=cmr12 scaled \magstep2

\centerline{\bf STANFORD UNIVERSITY}
\centerline{\bf DEPARTMENT OF STATISTICS}
\centerline{\big DEPARTMENTAL SEMINAR}
\bigskip

\baselineskip=12pt
\centerline{4:15 p.m., Tuesday, November 2, 1999}
\centerline{Sequoia Hall Rm. 200}
\centerline{(Cookies at 3:45 in 1st Floor Lounge)}

\bigskip
\baselineskip=15pt
\centerline{\sl Trevor Hastie \& Rob Tibshirani}
\centerline{\sl Stanford University}
\bigskip
\centerline{\bf New Statistical methods for DNA microarrays}
\centerline{\bf }
\bigskip

It is now possible to simultaneously measure the expression of thousands
of genes during cellular differentiation and response, through the use of
DNA microarrays.  A major statistical task is to understand the structure
of the data that arise from this technology. A typical data set consists
of a few thousand gene expression measurements on each of about 50
samples.  Hierarchical clustering has proven to useful for describing the
groupings of different samples and different genes.  In this talk we
describe a new method called "gene shaving" which tries to find clusters
of genes that show large variation over the samples.  The technique can be
unsupervised, that is, treat the samples as unlabelled, or partially or
fully supervised by known class labels for the samples.  We illustrate the
technique on some cancer tumor data.

This work is joint with David Botstein (Genetics, Stanford), Pat Brown
(Biochemistry, Stanford) and Mike Eisen (Berkeley).

\bye