When: September 9, 2015, 2:00 PM
Location: 3rd Floor Orchard View Room , Discovery Building
Contact: 608-316-4401, firstname.lastname@example.org
Finding the interactions in a system using observational data without exhaustive computational testing
An important task in systems biology is identifying the key components of a system. This task can be difficult if components interact in complex ways to produce the measured outputs of the system. The classic example of a challenging interaction is the non-linear exclusive-or (XOR) of two components to produce a measured output. This talk presents a combination of non-linear support vector machines with a novel feature selection algorithm in order to identify the relevant components of a system. This novel combination is evaluated both on synthetic data, where the ground truth components are known and participate in XOR relationships or other such challenging “correlation-immune” relationships, and on SNP data from a region of the human genome believed to contain interactions that modulate breast cancer risk.
This is joint work with Niko Escanilla, Ross Kleiman, and James Shull
All SysBioM (Systems Biology in Madison) sponsored talks take place on Wednesdays at 2 p.m. in the 3rd floor Orchard View room of the Discovery Building. Talks are open to the public. Access to the room is via the elevator behind Aldo’s Cafe in the Northeast corner of the building.