Personal tools
You are here: Home Events ANC Workshop: Katharina Heil and Amos Storkey, Chair: Gavin Gray

ANC Workshop: Katharina Heil and Amos Storkey, Chair: Gavin Gray

— filed under:

  • ANC Workshop Talk
When Jan 19, 2016
from 11:00 AM to 12:00 PM
Where IF Room 4.31/4.33
Add event to calendar vCal

Katharina Heil

Systems Biological approach to Parkinson's Disease *Uncovering the effects of PD on the synapse*

Traditional models of Parkinson?s Disease (PD) focus on events at the neuronal synapse. Although recently, non-synaptic pathways have been linked to PD, most therapies still target the synapse. Large-scale Genome Wide Association Studies (GWAS), mRNA array analyses and neuronal proteomics give new insights into causal and pathological mechanisms at all levels: genes - transcripts - proteins. Here we examined correlations amongst different datasets, specifically looking for evidence of synaptic mechanisms involved in PD.

 A wide range of cellular compartments and pathways (e.g. mitochondria, ubiquitin pathway, synapses) have been implicated in PD. At the molecular/genetic level GWAS and microarray studies implicate ~550 and

~700 genes respectively. Additionally large scale neuroproteomics studies supply details of molecular complexes in neurons. Systems biology combines this multi-level information and broadens the understanding of complex patterns, using techniques to analyse overlapping sets of affected genes, their interaction patterns and co-occurrence under different conditions.

To gain further understanding, we obtained synaptic protein data and molecular interactions to reconstruct protein-protein interaction networks. These were clustered to identify networks and structural sub-clusters. We annotated PD associated genes on these networks as defined by literature review, GWAS reporting and microarray study results. Functional enrichment analyses were performed for the network clusters and their substructures, confirming significance through hypergeometric testing.

Thereby we found that molecules associated with PD via GWAS or microarray experiments (~1,200) rarely overlap indicating that they reveal different PD mechanisms. Their overlap with the synaptic proteome (~1,000 genes) is a relatively small fraction of implicated molecules (1/6th of PD-associated genes), confirming the implication of further pathways in PD. We observed specific PD enriched subclusters at the synapse associated with vesicle cycling and protein folding.

 Our analysis confirms an emerging picture that causal mechanisms and pathological responses in PD are certainly synapse related but comprise a relatively small part of the overall picture. Subcomplexes in pre- and post-synapse were discovered as potentially involved, differing in response to the disease cause (GWAS) and pathology (array).


Amos Storkey

A whistle-stop tour of transactional learning.

I will review the results of the last few years in transactional machine learning. This will move from choices of optimal aggregators, through a variety of crowdsourcing incentive mechanisms, privacy in trading information, to incentivised collaboration over latent representations.

Transactional machine learning maps standard machine learning problems into problems that many people can collaborate to solve. By ensuring explicit incentive mechanisms for each individual, ideal collaborative outcomes can simply be achieved by each player doing their best on their local objective. In the end, machine learning is just a game.