Property testing and its connection to learning and approximation
Journal of the ACM (JACM)
On the approximability of numerical taxonomy (fitting distances by tree metrics)
Proceedings of the seventh annual ACM-SIAM symposium on Discrete algorithms
New algorithms for the duplication-loss model
RECOMB '00 Proceedings of the fourth annual international conference on Computational molecular biology
Robust Characterizations of Polynomials withApplications to Program Testing
SIAM Journal on Computing
| Hi-index | 0.00 |
Evidence indicates that members of many gene families in the genome of an organism tend to have homologues both within their own genome and in the genomes of other organisms. Amongst these homologues, typically only one or a few per genome perform an analogous function in their genome. Finding subsets of these genes which show evidence of performing a common function is an important first step towards, for instance, the creation of phylogenetic trees, multiple sequence alignments and secondary structure predictions. Given a collection of taxa P = {P1, P2, ..., Pk} where Pi contains genes {pi, 1, pi, 2, ..., pi, ni}, we ask to choose one gene from each of the taxa Pi such that these chosen vertices most agree. We define most agreeing in three distinct ways: most tree-like, pairwise closest, and pairwise most similar. We show these problems to be computationally hard from almost every angle via classical, parameterized and approximation complexity theory. However, on the positive side, we give randomized approximation algorithms following ideas from [GGR98] for the pairwise closest and pairwise most similar variants.