Date: 02.06.2016, Thursday
Time: 12:30-14:00
Place: Room M2, Quantum

Speakers:

Bogdan Iancu
Emilia Kozlowska

Abstract:

In this computational-experimental joint work , the aim is to infer novel network connections at molecular level from global phosphoproteomics measurements, by making use of a previously developed method based on the responses to perturbations at steady state, called Modular Response Analysis (MRA). The method disentangles the direction and the connection strength between two components of a network by perturbing the network and quantifying how varying the response of a node affects the response of the other nodes in the system. However, it requires previous knowledge about the interaction topology of the underlying network. To overcome this practical challenge, we use a modified method which integrates MRA with Bayesian variable selection, and is particularly designed for inferring interactions in molecular networks from incomplete perturbation data. Using several simulated and real data examples, we investigate the performance and limitations of this method under various experimental setups, related to the network size, perturbation types, and measurement noise.

Date: 02.06.2016, Thursday
Time: 12:30-14:00
Place: Room M2, Quantum

Speakers:

Bogdan Iancu
Emilia Kozlowska

Abstract:

In this computational-experimental joint work , the aim is to infer novel network connections at molecular level from global phosphoproteomics measurements, by making use of a previously developed method based on the responses to perturbations at steady state, called Modular Response Analysis (MRA). The method disentangles the direction and the connection strength between two components of a network by perturbing the network and quantifying how varying the response of a node affects the response of the other nodes in the system. However, it requires previous knowledge about the interaction topology of the underlying network. To overcome this practical challenge, we use a modified method which integrates MRA with Bayesian variable selection, and is particularly designed for inferring interactions in molecular networks from incomplete perturbation data. Using several simulated and real data examples, we investigate the performance and limitations of this method under various experimental setups, related to the network size, perturbation types, and measurement noise.