A new maximum-likelihood approach to chromosome physical-mapping

map of a chromosome from clone-clone hybridization experiments for the
protocol used at the University of Georgia to map the fungus Aspergillus
nidulans as part of the Fungal Genome Initiative.

In this protocol, random intervals of the chromosome called "clones" are
extracted and selected to be of approximately the same length, and a maximal
non-overlapping subset of the clones, called the "probes", is known.
The output of the protocol is a 0-1 matrix summarizing the results of
"hybridization experiments" between the probe subset and the remaining clones,
which records whether a probe overlaps a clone and may contain false positives
and negatives. The goal is to reconstruct the order and spacing of the probes
from the hybridization matrix to obtain a minimal set of clones that cover the
chromosome.

We derive a likelihood function on probe orders and spacings under a natural
probability model and describe a procedure that seeks to find the
maximum-likelihood map for this function by a combination of continuous
and discrete optimization. The new approach has been successfully applied
to chromosome VI of A. nidulans.

Joint work with Sanjay Shete and Jonathan Arnold of the Departments of
Statistics and Genetics at the University of Georgia.