References BAR+ Annotation Community detection

BAR+ (Piovesan et al. 2011) is a non hierarchical clustering method relying on a non comparative large-scale genome analysis. The present version of BAR+ contains 913,762 clusters with over 9 million sequences; in 10 % of the clusters, including some 5 million sequences, structural and functional features are statistically validated. The core idea of BAR+ is that when a sequence sharing at least 40 % sequence identity over at least 90 % of the alignment length with one of the sequence in a validated cluster it inherits structural and functional annotations from the cluster. Sequences in a cluster inherit annotations from proteins that have been experimentally characterised, when the feature/s is/are statistically meaningful (P-value < 0.01) after evaluating the cumulative distribution of Bonferroni corrected P-values (Bartoli et al. 2009). Features may include GO terms of the three different branches (Molecular Function, Biologica