Showing papers presented at "International Conference on Bioinformatics in 1994"

Automated integration of genomic physical mapping data via parallel simulated annealing

Abstract: The Human Genome Center at the Lawrence Livermore National Laboratory (LLNL) is nearing closure on a high-resolution physical map of human chromosome 19 We have build automated tools to assemble 15,000 fingerprinted cosmid clones into 800 contigs with minimal spanning paths identified These islands are being ordered, oriented, and spanned by a variety of other techniques including: Fluorescence Insitu Hybridization (FISH) at 3 levels of resolution, ECO restriction fragment mapping across all contigs, and a multitude of different hybridization and PCR techniques to link cosmid, YAC, AC, PAC, and Pl clones The FISH data provide us with partial order and distance data as well as orientation We made the observation that map builders need a much rougher presentation of data than do map readers; the former wish to see raw data since these can expose errors or interesting biology We further noted that by ignoring our length and distance data we could simplify our problem into one that could be readily attacked with optimization techniques The data integration problem could then be seen as an M x N ordering of our N cosmid clones which ``intersect`` M larger objects by defining ``intersection`` to mean either contig/map membership or hybridization results Clearly, the goal of making an integrated map is now to rearrange the N cosmid clone ``columns`` such that the number of gaps on the object ``rows`` are minimized Our FISH partially-ordered cosmid clones provide us with a set of constraints that cannot be violated by the rearrangement process We solved the optimization problem via simulated annealing performed on a network of 40+ Unix machines in parallel, using a server/client model built on explicit socket calls For current maps we can create a map in about 4 hours on the parallel net versus 4+ days on a single workstation Our biologists are now using this software on a daily basis to guide their efforts toward final closure