Description

Prof. Saman K. Halgamuge FIII, The University of Melbourne delivered a talk on "AI and Learning Algorithms for Biomedical Engineering in Recent PhD Research Projects" on 20th Devember 2018 as a part of the BME seminar series organized by the IEEE EMBS Student Branch Chapter of University of Moratuwa. Abstract of the Talk : Deep Learning. In particular the Unsupervised Deep Learning and various methods of exploratory data analytics have been extensively used Bioinformatics In particular in Metagenomics but also increasingly in other areas of knowledge discovery Including Neural Engineering. The direct interaction between drugs, drug characterizations using data collected from Multiple Electrode Arrays (MEAD) and the computational work on drug repositioning have been the major focus of Bioinformatics collaboration between our research group and Howard Florey Research Institute at University of Melbourne. Repositioning of existing drugs as appropriate medication for previously not associated medical conditions help reduce the time, costs and risks of drug development. Identification of drug groups either as clusters or subnetworks has already been used to simplify the visualization and Interpretation of data for the purpose of drug repositioning. MEA Is an extracellular recording technology that enables the analysis of networks of neurons in vitro by producing "big data". Neurons in culture exhibit a range of behavioral dynamics, which can be measured in terms of Individual action potentials, network-wide synchronized firing and a host of other features that characterize network activity. MEA data analysis is used to differentiate between two types of antiepileptic drugs with different mechanisms of action. It Initially extracts features that characterize different aspects of neuronal activity that can be used to characterize network states. This utilizes existing feature extraction methods as well as novel methods that are adaptive to activity patterns in unperturbed and perturbed network states. These features are then used to build network signatures that allow novel compounds to be compared with compounds with known mechanisms of action. This research demonstrates that MEA-based workflows can assist in rapid and efficient screening of pharmacological compounds, making them a useful addition to drug development pipelines.