Computational Biology
Regulatory Network Modeling
We develop predictive models of gene regulation. Our algorithms empirically model the relationship between regulators and target genes, using high throughput data. These efforts include constrained regression model (Galagan 2013) and probabilistic frameworks such as module networks (Azizi 2014).
The frameworks are designed to be flexible in modeling complex regulatory effect while maintaining tractability. The models provide system-level understanding of the organism, providing answers to biomedical problems concerning the organism and insights to similar biological systems.
Metabolic Network Modeling
We develop predictive models of metabolism based on Flux-Balance Analysis. While complete genome sequences describe the range of metabolic reactions that are possible for an organism, they cannot quantitatively describe the behaviour of these reactions. We developed the first algorithm - E-flux - for integrating gene expression into FBA to predict metabolic state. E-flux can be used to predict which drugs specifically target selected pathways (Colijn 2009), determine the nutrients being utilized by an organisms (Brandes 2012), and provide insight into the metabolic functions of transcriptional regulators (Garay 2015).
Data Analysis Algorithms
We develop algorithms to be part of our pipeline to process high throughput genomics data. For example, our ChIP-Seq analysis tool BRACIL (Gomes 2014), based on the use of blind deconvolution for signal detection (Lun 2009), identifies binding sites with single-nucleotide resolution from the enriched region and detects cooperative interactions when the TF binds two sites simultaneously.