James Owers and Douglas Armstrong
Tue 06 Feb 2018, 11:00 - 12:00
IF 4.31/4.33

If you have a question about this talk, please contact: Gareth Beedham (gbeedham)

James Owers

Title: Can we learn a musical instrument synthesizer?

Abstract:  Beginning with a (very) brief history of analogue synthesis of signals, I’ll discuss the feasibility of learning an synthesizer which emulates the sounds of realistic musical instruments. Additionally, I’ll explore some applications of machine learning to music which could result in tools for musicians.


Douglas Armstrong

Title: Network topology analysis reveals substructures within the synaptic proteome that have close association with complex traits.

Abstract: The synaptic proteome is widely believed to be the molecular machine that underpins the core functions of neurons – the integration and transfer of information from one cell to another. Perhaps unsurprisingly it is enriched for proteins whose genes are linked to a wide range of human neurological conditions. However GWAS datasets from these conditions map less clearly onto the synaptic proteome often with weak or no significant enrichment. The synaptic proteome can be subdivided, on the basis of network topology  into cluster that each have enriched functional associations. We hypothesised that these topological communities form natural groups for gene set analysis and contain information not only about protein encoding genes with a mechanistic association with the phenotype but also with interaction partners whose role is important but less obvious. We combined 30 published synaptic proteomic studies from 2000 to date to obtain a list of 6500 molecules. We retrieved protein-protein interactions (PPIs) for combined list and built the most complete up-to-date PPI networks for presynaptic and postsynaptic compartments. We then divided this proteome up into sub-communities on the basis of network topology. We analysed three large Genome Wide Associations Studies of genetic associations with human cognitive ability or educational attainment. A single community was found to have a significantly enriched association with educational attainment (n>100,000) and this was then replicated across the other cohorts. The community contained half of the synaptic glutamate receptors. However, the glutamate receptors (as a group) are under strong evolutionary selection pressure and showed no significant association with the phenotype on Genome Wide Association analysis (GWAS). The majority of the signal in the community arose from other proteins in a dense subgraph closely linked to the glutamate receptors.