Evidence for intelligent (algorithm) design

More than 700 computational biologists convened in beautiful Venice in early April for RECOMB 2006, the 10th annual Conference on Research in Computational Molecular Biology. After 40 talks, 6 keynote lectures, 180 posters, and at least two cameos by the Riemann zeta function, several emerging trends in computational biology are apparent.First, there has been a strong shift towards empirical studies of molecular evolution and variation, with approximately 25% of the papers in this broad area. We expect that this number can only increase in the near future, given the ENCODE project http://www.genome.gov/10005107 webcite and the forthcoming release of several new eukaryotic genomes.Second, there is a resurgence of interest in two of the oldest problems in computational biology: RNA folding and protein sequence alignment. The interest in noncoding RNAs (ncRNAs) is driven by experiment: recent work on RNA interference (RNAi), microRNAs, ribozymes, and the rest of the 'modern RNA world' has once again stimulated interest in the classical problems of ncRNA identification and fold prediction. Advances in protein sequence alignment draw on the development of new algorithmic and machine-learning techniques for principled estimation of gap penalties (the penalty for inserting a gap in the alignment to improve it) and the rigorous incorporation of non-local similarity measures that move beyond residue-residue similarity.Interest in classic areas such as protein structure and folding remains strong, with several papers tacitly or explicitly motivated by the coming flood of data promised by structural genomics. Many of the other mainstays of computational biology were also represented at the conference, including old favorites such as expression analysis and genome evolution, as well as the newer areas of data integration and network alignment. Notable by their absence were papers on genome assembly and human single-nucleotide polymorphism (SNP) variation; this is likely to be a