Bayesian Multiple-Instance Motif Discovery with BAMBI
― Inference of Recombinase and Transcription Factor Binding Sites

― Reconstruction of Novel Transcription Factor Regulons


INTRODUCTION - PUBLICATIONS - TOOLS - ACKNOWLEDGMENTS


 

Overview

Results

Ongoing

Principal Contributors

 

 

Overview

 

Finding conserved motifs in genomic sequences represents one of essential bioinformatic problems. However, achieving high discovery performance without imposing substantial auxiliary constraints on possible motif features remains a key algorithmic challenge.
   

This website provides an overview of BAMBI ― Bayesian Algorithm for Multiple Biological Instances of motif discovery ― a sequential Monte Carlo motif-identification algorithm. BAMBI is based on a position weight matrix (PWM) model that does not require additional constraints and is able to estimate such motif properties as length, logo, number of instances, and their locations solely on the basis of primary nucleotide sequence data. Furthermore, should biologically-meaningful information about motif attributes be available, BAMBI takes advantage of this knowledge to further refine the discovery results.
[TOP]

 

Results

 

In practical applications, we have shown that the proposed approach can be used to find binding sites of the cAMP receptor protein (CRP) and of Din-family site-specific serine recombinases. Published results obtained in these and other settings demonstrate that BAMBI evinces better statistical performance ― as measured by the nucleotide-level correlation coefficient (nCC) ― than several of the widely-used profile-based motif discovery methods, including: MEME, BioProspector with BioOptimizer, SeSiMCMC, and Motif Sampler. Additionally, in the case of Din-family recombinase target site discovery, the BAMBI-inferred motif was found to be the only one functionally accurate from the underlying biochemical mechanism standpoint.

[TOP]

 

Ongoing

 

Recent progress includes the development of the next generation algorithm, BAMBI-2b, which incorporates numerous technical and methodological improvements, including the ability to recognize and utilize the "2-block" structures present in many DNA binding sites as well as the potential (inverted) repeat symmetries of these blocks.

 

The extended BAMBI-2b algorithm is currently being applied to the problem of transcription factor binding site discovery and their regulon identification. To date, BAMBI-2b performance has been successfully tested on several known E. coli transcription factors and their regulons. Ongoing work involves applying BAMBI-2b to the expansion of known regulons, including their enrichment with intra-genic and distant upstream sites, as well as the identification of novel regulons in species where they have not been fully annotated ― particularly focusing on D. alaskensis G20.

[TOP]

 

Principal Contributors

Abdulkadir Elmas

Electrical Engineering Department, Columbia University

 

Michael S. Samoilov

QB3 California Institute for Quantitative Biosciences, University of California Berkeley

 

Guido H. Jajamovich

Electrical Engineering Department, Columbia University

 

Xiaodong Wang

Electrical Engineering Department, Columbia University

 

Adam P. Arkin

QB3 California Institute for Quantitative Biosciences, University of California Berkeley

Department of Bioengineering, University of California Berkeley, Center for Synthetic Biology,

Physical Biosciences Division, Lawrence Berkeley National Laboratory


[TOP]