Validation of computational prediction of horizontal gene transfer events—XenoCluster

Abstract

In this paper, we assess the biological accuracy and relevance of XenoCluster results compared to a reported set of known xenologs in yeast (Hall et al. Eukaryot. Cell 4(6):1102---1115, 2005). We were able to assign a high-priority rank to all reported yeast xenologs for which we had sufficient genomic sequence data. Additionally, we now report on a fine-grained extension to our previous implementation (Walters et al. 8th International Parallel Computing Technologies Conference (PaCT), LNCS, vol. 3606, pp. 355---366, Springer, Berlin, 2005), in which we achieved 80% efficiency on 8 CPU cores. In the coarse-grained component, high degrees parallelism (up to 65x speedup on 4,096 processors) was reported. While an obvious candidate for coarse-grained implementation, this paper now describes a multiple granularity parallelism solution that includes exploitation of multicore shared memory nodes to address fine-grained aspects in the tree-clustering phase of our previous deployment of XenoCluster (Walters et al. 8th International Parallel Computing Technologies Conference (PaCT), LNCS, vol. 3606, pp. 355---366, Springer, Berlin, 2005).