MITSU

MITSU discovers transcription factor binding site (TFBS) motifs by integrating stochastic Expectation-Maximization (sEM) with an improved likelihood approximation to enhance motif discovery accuracy.

Key Features:

• Algorithmic Innovation: Combines stochastic Expectation-Maximization (sEM) with an enhanced approximation to the likelihood function for unconstrained modeling of motif occurrences.
• Performance Superiority: Evaluated on synthetic data and characterized prokaryotic TFBS motifs, showing superior site-level positive predictive value compared to standard EM and alternative sEM-based algorithms.
• Implementation: Packaged as a Java executable with compatibility for Linux and OS X environments.

Scientific Applications:

• Gene regulation analysis: Identification of TFBS motifs to support studies of transcriptional regulatory networks and their roles in cellular processes.
• Prokaryotic motif characterization: Validation and discovery of motifs in prokaryotic systems using characterized TFBS datasets.

Methodology:

MITSU uses iterative sampling and updating procedures within a stochastic EM (sEM) framework, incorporates an improved likelihood approximation, and models motif occurrences without positional constraints to estimate TFBS motifs.