Phylogibbs

Phylogibbs identifies regulatory binding sites in collections of DNA sequences, including multiple alignments of orthologous intergenic regions from related organisms, for de novo discovery of transcription factor motifs.

Key Features:

• Phylogenetic consideration: Explicitly models phylogenetic relationships between species to distinguish functional conservation of binding sites from conservation due to evolutionary proximity.
• Multiple sequence alignment support: Operates on arbitrary collections of multiple local sequence alignments of orthologous sequences and explores configurations for assigning binding sites across these alignments.
• Multi-factor capability: Accommodates an arbitrary number of transcription factors (TFs) when assigning motifs and sites.
• Bayesian probabilistic scoring: Uses a Bayesian probabilistic model to score binding site configurations accounting for the evolution of binding sites and background intergenic DNA.
• Simulated annealing and MCMC sampling: Applies simulated annealing combined with Monte Carlo Markov-chain (MCMC) sampling to assign posterior probabilities to reported binding sites.

Scientific Applications:

• Performance on Saccharomyces data: Demonstrated superior motif-recovery performance on synthetic and real data from five Saccharomyces species, recovering over 50% of binding sites in S. cerevisiae at ~50% specificity and 33% at ~85% specificity.
• ChIP-on-chip integration: Identified motifs for 16 out of 21 transcription factors when applied to collections of multiple alignments annotated with ChIP-on-chip data.
• Literature concordance and target validation: Produced predictions that match literature consensus in most cases of algorithmic disagreement and can recover known target genes from the literature.

Methodology:

Performs comparative analysis of orthologous intergenic regions using explicit phylogenetic modeling on multiple local sequence alignments, scores binding site configurations with a Bayesian probabilistic model that accounts for binding-site evolution and background DNA, and employs simulated annealing combined with MCMC sampling to obtain posterior probabilities for reported sites.