Overview

The focus of this research is to identify calmodulin binding sites within known calmodulin binding partners and also to identify novel interaction partners using binary and structured support vector machines.

Background

Calmodulin is a highly-conserved protein among eukaryotes. While the similarity among homologs is high, interaction partners and binding sites share little sequence conservation. However, as shown above, calmodulin binding sites tend to share some features, namely they are contiguous in sequence, 20 residues on average, and tend to form helices.

Methods

Binary and structural support vector machines are trained on a variety of kernels defined for proteins that capture sequence composition, physico-chemical properties, and sequence conservation. See my thesis [1] and extended MLG abstract [2] for more details. Predictions of calmodulin binding proteins in Arabidopsis can be found here.

References

Overview

This project aims to identify regulatory signals and create predictive models for alternative splicing in the plant Arabidopsis in collaboration with A.S.N. Reddy.