Algorithms for computing and integrating physical maps using unique probes
- Univ. of Arizona, Tucson, AZ (United States)
Current physical mapping projects based on STS-probes involve additional clues such as the fact that some probes are anchored to a known map and that others come from the ends of clones. Because of the disparate combinatorial contributions of these varied data items, it is difficult to design a {open_quotes}tailored{close_quotes} algorithm that incorporates them all. Moreover, it is inevitable that new experiments will provide new kinds of data, making obsolesent any such algorithm. We show how to convert the physical mapping problem into a 0/1 linear programming (LP) problem. We further show how one can incorporate additional clues as additional constraints in the LP formulation. We give a simple relaxation of the 0/1 LP problem that solves problems of the same scale as previously reported tailored algorithms to equal or greater optimization levels. We also present a theorem proving that when the data is 100% accurate, then the relaxed and integer solutions coincide. The LP algorithm suffices to solve problems on the order of 80-100 probes - the typical size of the 2- or 3-connected contigs. Unlike previous work, this algorithm only links and orders contigs when the join is 90% or more likely to be correct. It is our view that there that there is no value in computing an optimal solution with respect to some criteria over very noisy data as this optimal solution rarely corresponds to the true solution. The paper involves extensive empirical trials over real and simulated. 11 refs., 9 figs., 3 tabs.
- Research Organization:
- Association for Computing Machinery, New York, NY (United States); Sloan (Alfred P.) Foundation, New York, NY (United States)
- OSTI ID:
- 549006
- Report Number(s):
- CONF-970137--; CNN: Grant LM-04960
- Country of Publication:
- United States
- Language:
- English
Similar Records
A Quantum-Inspired Tensor Network Algorithm for Constrained Combinatorial Optimization Problems
Certifiably Correct Range-Aided SLAM