Methods

With biological inputs varying from cell lines to patient-derived circulating tumour cells, our integrated Systems Microscopy pipeline derives mechanistic and diagnostic insights, as well as enabling high-throughput drug / therapy discovery. This approach supports both fundamental and clinical research through efficient (re)use of a core set of experimental and analytical tools.

Australia’s only dedicated Systems Microscopy pipeline

Our Systems Microscopy approach integrates multidisciplinary methods into a coherent pipeline. Central to this pipeline are automated experiments using liquid handling systems to enhance throughput and reproducibility. This is followed by automated fluorescence imaging using dedicated microscopy hardware (confocal and epifluorescence), with “Proteomic Microscopy” a major new capability unlocking the potential of in vivo, spatially resolved systems biology (see ‘1’ below). After image acquisition, automated analysis (see ‘2’ below) is performed using open source software platforms, as well as tools developed locally with machine vision experts.

These tools extract objective, high-dimensional quantitative data from biological images, facilitating downstream statistical, machine learning and deep learning analyses (see ‘3’ below). We are developing unique data visualization tools to coherently assess and interpret the resulting wealth of data, using immersive environments, traditional desktop and cutting-edge VR platforms. This work, in collaboration with the Expanded Perception and Interaction Centre (EPICentre, UNSW Art & Design) is providing capacity for interpretation and collaboration - for cellular data, as well as analyses of clinical / biomedical images (X-ray, MRI, CT-scan etc).

1) Automated cellular imaging & Proteomic Microscopy

Automation of experimental and imaging protocols supports reproducible, objective and high-throughput research methods. We integrate live and fixed (correlative) imaging to extend molecular analysis of dynamic processes. With support from the Ramaciotti Systems Microscopy Unit of the UNSW Biomedical Imaging Facility, we are deploying highly multiplexed fluorescence imaging of 20+ molecular components in each individual cell. This Proteomic Microscopy provides detailed insights into the composition and spatial organisation of single cells and sub-cellular compartments.

2) Quantitative image analysis including machine vision

A variety of software is used to analyse fixed and dynamic cellular or tissue image data at the single cell and sub-cellular object level. Predominantly using open source tools, we also collaborate with experts in machine vision and deep learning to optimise cell / object detection and classification methods. Automated image analysis is a pivotal aspect of our work, converting visual information into quantitative data facilitating objective statistical and machine learning analyses generating new biological and clinical insights.

3) Statistical analysis, AI & immersive data visualization

We employ quantitative analysis and visualization tools to translate biological information into insights. In-house, we use Knime as a backbone for rapid prototyping of automated analysis pipelines, with R for more challenging problems involving multivariate statistics, dimension reduction and manifold projection, dynamic lineage inference, information theoretic analyses of causal relationships and regulatory network structure. Machine learning approaches guide feature selection (e.g. which biological features are important to define a state or process) and to develop predictive models. We work with experts in AI-driven analysis and large-scale data visualization to support enhanced interpretation, including through development of unique immersive- and VR-based data visualization software (BioDive 2.0 - see below!).

BioDive 2.0 - Virtual Reality-based Visual Analytics Software for immersive exploration of quantitative Single Cell Image Data

BioDive Software was developed using UNITY for use in large format immersive 3D visualisation systems. We have updated this software with a range of new capabilities, producing BioDive 2.0, which is now accessible on any screen and in virtual reality (VR). VR provides the largest possible visual ‘real estate’ for observing single cell images embedded in low dimensional projections of the high dimensional state space heterogeneity, based on our quantitative image analysis methods (above). This video (right) depicts some of the many different ways we can now interact with and explore images and quantitative data in an intuitively integrated environment.